Loading…
Ray Summit 2020 has ended
View More Details for Ray Summit & Registration Information.
Please note: All Sessions are in Pacific Daylight Time (PDT), UTC-7

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Tuesday, September 29
 

10:00am PDT

Anyscale Academy (Pre-Registration Required)
Register for Anyscale Academy.


Fee:
Attendees: $100
Students: $50


This five hour tutorial will occur in two sessions, 2.5 hours in the morning and 2.5 hours in the afternoon, will take you from being a Ray beginner to using Ray for reinforcement learning. You will receive a certificate of completion from Anyscale after completing the course and taking the two quizzes in the last 30 minutes of each session.

First Session | Ray Crash Course and Intro to Reinforcement Learning
September 29 | 10:00 AM – 12:30 PM, Pacific Daylight Time (PDT), UTC -7 
Ray was created to make it easier to scale diverse computation tasks and distributed state across a cluster, with a minimum of distributed systems expertise and knowledge required. You’ll learn how Ray meets these goals with a concise, intuitive API while performing efficient scheduling and execution of tasks for you. You'll learn this API and see how it breaks through the constraints of the Python interpreter's global interpreter lock. Whether you need better utilization of the cores in your workstation or you need massive compute resources scheduled across a cluster, you'll understand how to leverage Ray to meet your computation needs.

Then you’ll learn the core concepts of reinforcement learning that we’ll explore in greater depth in the afternoon. This morning tutorial provides foundational “building blocks” for Ray in general and use of RL in particular. We’ll use those building blocks for the afternoon tutorial.


Lessons:
  • Ray Tasks: Distributed, stateless computing
  • Ray Actors: Distributed, stateful computing
  • Ray Multiprocessing: Ray replacements for popular multiprocessing and multithreading libraries that let you break the single-node boundary.
  • Introduction to Reinforcement Learning: Learning core concepts of RL while solving a popular test environment (CartPole) with production-ready algorithms and tools.
  • Quiz - 60-70% pass required for the certificate


Second Session | Ray RLlib for Reinforcement Learning: Multi-armed Bandits and Recommendation Systems
September 29 | 2:00 PM – 4:30 PM Pacific Daylight Time (PDT), UTC -7
Using hands-on examples, we'll learn how to use RLlib and Tune to train and run reinforcement learning systems. This session builds on the morning’s introduction to reinforcement learning concepts.


Lessons:
  • Optimizing Market Investments with Multi-Armed Bandit: A real-world problem addressed with a “constrained” class of RL algorithms.
  • Keystone lesson: RL for Recommender Systems: New approaches to recommenders, which can be adapted to similar use cases, such as personalization.
  • Quiz - 60-70% pass required for the certificate


Tuesday September 29, 2020 10:00am - 4:30pm PDT
 
Wednesday, September 30
 

9:00am PDT

Welcome & Keynotes Introduction
Wednesday September 30, 2020 9:00am - 9:05am PDT
Virtual

9:00am PDT

Keynote: The Future of Ray - Robert Nishihara, CEO & Co-founder, Anyscale



Speakers
avatar for Robert Nishihara

Robert Nishihara

Co-founder and CEO, Anyscale


Wednesday September 30, 2020 9:00am - 9:15am PDT
Virtual

9:15am PDT

Keynote: Programming the Cloud as Easily as your Laptop - Ion Stoica, Executive Chairman, Anyscale | Professor, University of California Berkeley
In this talk, I will discuss the trends in computing and how these trends are accelerating the transition to distributed computing. Then, I will discuss the enormous challenges faced by developers in building distributed applications today, and how Anyscale and Ray address these challenges.


Speakers
avatar for Ion Stoica

Ion Stoica

Executive Chairman, Anyscale | Professor, University of California Berkeley


Wednesday September 30, 2020 9:15am - 9:30am PDT
Virtual

9:30am PDT

Keynote: Anyscale Product Demo - Edward Oakes, Software Engineer, Anyscale
This demo will illustrate how the Anyscale Platform enables you to build an end-to-end distributed AI application in a matter of minutes.

Speakers
avatar for Edward Oakes

Edward Oakes

Software Engineer, Anyscale
Edward Oakes is a software engineer at Anyscale and a graduate student at U.C. Berkeley.


Wednesday September 30, 2020 9:30am - 10:00am PDT
Virtual

10:05am PDT

Keynote: Building a Fusion Engine with Ray - Dr. Charles He, Chief Architect of Storage and Compute, Ant Group
During the course of serving 1.3 billion users on Alipay, Ant Group found itself living with a multitude of computing paradigms (such as AI, graph and streaming) This brought about challenges in performance and development efficiency. In this keynote Charles will talk about why and how Ant built a fusion engine with Ray to break computing boundaries.

Speakers
avatar for Charles He

Charles He

Chief Architect, Ant Group
Charles He, Ant Group, Chief Architect of cCmpute and Storage. Charles Changhua He is currently chief architect of compute and storage at Ant Group, operator of Alipay, world’s largest digital payment platform. He obtained PhD in computer science from Stanford University and served... Read More →


Wednesday September 30, 2020 10:05am - 10:15am PDT

10:15am PDT

Keynote: Scalable Python We Can Afford - Wes McKinney, Director of Ursa Labs, Creator of Python Pandas Project
In this talk, I will discuss some of the computational systems challenges that go hand in hand with doing scalable distributed computing in Python. In particular, I will explain how working to improve data processing performance and efficiency in data science tools led me to help build the Apache Arrow project.

Speakers
avatar for Wes McKinney

Wes McKinney

Director of Ursa Labs, Creator of Python Pandas Project, Ursa Labs | Python Pandas Project


Wednesday September 30, 2020 10:15am - 10:30am PDT
Virtual

10:30am PDT

Keynote: Easier Machine Learning Thoughts From Scikit-Learn - Gaël Varoquaux, Research Director, Inria
I will discuss how we have progressed on democratizing machine-learning, when building scikit-learn, tackling the bigger challenge of scalability, and developing new practices for machine learning on dirty data.

Speakers
avatar for Gaël Varoquaux

Gaël Varoquaux

Research Director, Inria
Gaël Varoquaux is a tenured research director at Inria. His research focuses on statistical-learning tools for data science and scientific inference. Since 2008, he has been exploring data-intensive approaches to understand brain function and mental health. He is one of the leaders... Read More →



Wednesday September 30, 2020 10:30am - 10:50am PDT
Virtual
  Keynote Sessions
  • Slides Included Yes

10:50am PDT

Stretching Break: Shoulders & Neck - Maryam Sharifzadeh, Office Yoga
Join us for 15 minutes of stretching.  No equipment required, do it right at your desk!

Wednesday September 30, 2020 10:50am - 11:05am PDT
Virtual

10:50am PDT

Break
Wednesday September 30, 2020 10:50am - 11:20am PDT
Virtual

10:55am PDT

Sponsored Office Hours: Anyscale
Anyscale was founded by the creators of Ray, an open source project from the UC Berkeley RISELab. Anyscale enables developers of all skill levels to easily build applications that run at any scale, from a laptop to a data center.

Wednesday September 30, 2020 10:55am - 11:25am PDT

11:20am PDT

Keynote: ML for Systems and Chip Design - Azalia Mirhoseini, Senior Research Scientist, Google Brain
In the past decade, computer systems and chips have played a key role in the success of AI. Our vision in Google Brain's ML for Systems team is to use AI to transform the way in which computer systems and chips are designed. Many core problems in systems and hardware design are combinatorial optimization or decision making tasks on large graphs. In this talk, we will describe some of our latest learning based approaches to tackling such large-scale optimization problems and demonstrate how AI agents can automate and optimize complex tasks such as chip floorplanning in 6 hours, whereas existing baselines require human experts in the loop and take several weeks.

Speakers
avatar for Azalia Mirhoseini

Azalia Mirhoseini

Senior Research Scientist, Google
Azalia Mirhoseini is a Senior Research Scientist at Google Brain and an Advisor at Cmorq. She is the co-founder/lead of the Machine Learning for Systems Moonshot at Brain where they focus on deep reinforcement learning based approaches to solve problems in computer systems and metalearning... Read More →


Wednesday September 30, 2020 11:20am - 11:40am PDT
Virtual

11:40am PDT

Keynote: Modern App Development - Adrian Cockcroft, VP Cloud Architecture Strategy, AWS
Speakers
avatar for Adrian Cockcroft

Adrian Cockcroft

VP Cloud Architecture Strategy, AWS
Adrian Cockcroft has had a long career working at the leading edge of technology, and is fascinated by what happens next. In his role at AWS, Cockcroft is focused on the needs of cloud native and “all-in” customers, and leads the AWS open source community development program.Prior... Read More →


Wednesday September 30, 2020 11:40am - 12:00pm PDT
Virtual

12:00pm PDT

Keynote: Artificial Innovation Accelerates Artificial Intelligence - David Patterson, Professor Emeritus, Professor in the Graduate School, UC Berkeley
Machine learning is now driving AI, and the first word in Machine Learning is machine.  Thus, we need even faster computers to enhance AI, but the slowing of Moore’s Law means conventional machines are barely improving. In order to  expand the impact of AI in the cloud and on the edge, we are forced to design machines specifically for AI.
 
Google is a leader on this new architectural path, with the first of three generations of TPUs deployed in the cloud since 2015. 
 
Machines for the edge also need to be tailored for AI. I’ll describe RISC-V, an architecture developed at the University of California, Berkeley designed to be easy to tailor on the edge that may become as popular for open source hardware as Linux is for open source software. I’ll also explain the role of the new RISC-V International Open Source (RIOS) Lab—based jointly in Berkeley and Shenzehn—in improving and expanding the RISC-V ecosystem.

Speakers
avatar for David Patterson

David Patterson

Professor Emeritus, Professor in the Graduate School, UC Berkeley
David Patterson is the Pardee Professor of Computer Science, Emeritus at the University of California at Berkeley, which he joined after graduating from UCLA in 1976.Dave's research style is to identify critical questions for the IT industry and gather inter-disciplinary groups of... Read More →


Wednesday September 30, 2020 12:00pm - 12:30pm PDT
Virtual

12:30pm PDT

Beach Party Break
Join us for some beach party tunes to give your mind a break, relax, and groove along to!

Wednesday September 30, 2020 12:30pm - 12:40pm PDT
Virtual

12:30pm PDT

Break
Wednesday September 30, 2020 12:30pm - 12:50pm PDT
Virtual

12:50pm PDT

Keynote: AI for Financial Services - Manuela Veloso, Head of AI Research, J.P. Morgan
In this talk, I will discuss AI in Finance, by briefly presenting a few research projects that we are carrying at JPMorgan AI Research.

Speakers
avatar for Manuela Veloso

Manuela Veloso

Head of AI Research, J.P. Morgan


Wednesday September 30, 2020 12:50pm - 1:10pm PDT
Virtual

1:10pm PDT

Keynote: Machine Learning & A.I. At Uber - Zoubin Ghahramani, Chief Scientist & VP, Artificial Intelligence, Uber Technologies
Machine learning and AI methods are at the heart of Uber's technology. I discuss areas where ML is applied at Uber, describe limitations of current methods, and provide some context on the frontiers of AI and ML research with a special focus on probabilistic AI and optimization.

Speakers
avatar for Zoubin Ghahramani

Zoubin Ghahramani

Chief Scientist & VP, Artificial Intelligence, Uber Technologies
Zoubin Ghahramani is Chief Scientist and VP of Artificial Intelligence at Uber Technologies, where he serves on the Engineering Leadership Team and leads the engineering and advanced research teams developing data-driven algorithms to optimize the services Uber provides users worldwide... Read More →


Wednesday September 30, 2020 1:10pm - 1:30pm PDT
Virtual

1:30pm PDT

Disco Party Break
Join us for some disco party tunes to give your mind a break, relax, and groove along to!

Wednesday September 30, 2020 1:30pm - 1:40pm PDT
Virtual

1:30pm PDT

Break
Wednesday September 30, 2020 1:30pm - 2:00pm PDT
Virtual

1:30pm PDT

Sponsored Office Hours: Arimo
H1st AI solves the “cold-start” problem of Industrial AI: encoding human expertise to augment the lack of data, while bridging to powerful ML—based on experience building AI solutions at Panasonic: robotics predictive maintenance, cold-chain energy optimization, Gigafactory battery manufacturing, avionics, automotive cybersecurity, and more.

Wednesday September 30, 2020 1:30pm - 2:00pm PDT

2:00pm PDT

Building a Python Web Service with Ray - Philipp Moritz, Anyscale
If you are building web services in Python that need to scale, this talk is for you! We show how the Python web serving ecosystem integrates with Ray and how Ray makes it easy to scale up web services and to manage their state. We address practical challenges such as how to use Ray tasks and actors effectively, how to write asynchronous code, and how to do type checking and testing. We also cover deployment challenges like monitoring and tracing in a distributed web service. The talk draws on lessons we learned from building Anyscale. We will dive into Anyscale’s architecture and the technology choices we made and how they evolved.

Speakers
avatar for Philipp Moritz

Philipp Moritz

Co-founder and CTO, Anyscale


Wednesday September 30, 2020 2:00pm - 2:30pm PDT
Virtual 1

2:00pm PDT

Hyperparameter Tuning and Visualization in Deep Learning - Lukas Biewald, Weights & Biases
Deep learning models are incredibly powerful but often tricky to adapt to new use cases. Whether you’re finetuning a pretrained net on new data, trying to build an intuition for a complex model, or throwing a variety of architectures at a unique problem, hyperparameter exploration can help. We will share high-level approaches and useful visualizations for hyperparameter search, grounded in concrete examples from semantic segmentation, image classification, language understanding, and other domains. Though we will focus on Weights & Biases Sweeps as a comprehensive tool for this task, these practices are framework-agnostic, and we hope they can accelerate your progress regardless of your dev setup.

Speakers
avatar for Lukas Biewald

Lukas Biewald

Founder, CEO, Weights & Biases


Wednesday September 30, 2020 2:00pm - 2:30pm PDT
Virtual 3

2:35pm PDT

Easy Access to SOTA NLP Models with Ray and Hugging Face - Thomas Wolf, Hugging Face
In this talk, I'll discuss how NLP researchers and practitioners can leverage Hugging Face models and datasets libraries together with Ray distributed tools to use and train the latest state-of-the-art NLP models.

Speakers
avatar for Thomas Wolf

Thomas Wolf

Co-founder and Chief Science Officer, Hugging Face
Thomas Wolf is co-founder and Chief Science Officer of Hugging Face. His team is on a mission to catalyze and democratize NLP research. Prior to HuggingFace, Thomas gained a Ph.D. in physics, and later a law degree. He worked as a physics researcher and a European Patent Attorney... Read More →



Wednesday September 30, 2020 2:35pm - 3:05pm PDT
Virtual 2

2:35pm PDT

Ray: A General Purpose Serverless Substrate? - Eric Liang, Anyscale
Everyone wishes their app was "serverless". You don't have to worry about machines, and it's very elastic, so you get the best of both worlds on performance and cost. However, the reality today is that many classes of applications are infeasible to build serverlessly today: either they don't fit into or cannot depend on proprietary serverless platforms and execution models. In this talk, we look at the key limitations these applications face, and a vision of how Ray 1.0 fits into the picture as a more general purpose serverless platform.

Speakers
avatar for Eric Liang

Eric Liang

Software Engineer, Anyscale
Eric Liang is a software engineer at Anyscale and a graduate student at U.C. Berkeley, where he does research on reinforcement learning.



Wednesday September 30, 2020 2:35pm - 3:05pm PDT
Virtual 1
  Ray and Its Libraries
  • Slides Included Yes

2:35pm PDT

EfficientBERT: Trading off Model Size and Performance - Meghana Ravikumar, SigOpt
With the publication of BERT, transfer learning was suddenly accessible for NLP, unlocking a plethora of model zoos and boosting performances for domain specific problems.  Although BERT has accelerated many modeling efforts, its size is limiting. In this talk, we will explore how to reduce the size of BERT while retaining its capacity in the context of English Question Answering tasks. We’ll show how scalable hyperparameter optimization can help you tackle difficult modeling problems, draw insights, and make informed decisions.

Our approach encompasses fine-tuning, distillation, architecture search, and hyperparameter optimization at scale. First, we fine-tune BERT on SQUAD 2.0 (our teacher model) and use distillation to compress fine-tuned BERT to a smaller model (our student model). Then, combining SigOpt and Ray, we use multimetric Bayesian optimization at scale to find the optimal architecture for the student model. Finally, we explore the trade-offs of our hyperparameter decisions to draw insights for our student model’s architecture.

Speakers
avatar for Meghana Ravikumar

Meghana Ravikumar

Machine Learning Engineer, SigOpt
Meghana has worked with machine learning in academia and in industry, and is happiest working on natural language processing. Prior to SigOpt, she worked in biotech, employing NLP to mine and classify biomedical literature. When she’s not reading papers, developing models/tools... Read More →



Wednesday September 30, 2020 2:35pm - 3:05pm PDT
Virtual 3
 
Thursday, October 1
 

9:00am PDT

AI in Production - Key Challenges & Use Cases - Jonathan Berte, Robovision
Scaling AI applications require a parallel pipeline approach. With the help of ray actors, Robovision created the concept of ML framework independent compute cells and pipelines. These cells can be reconfigured later into other pipelines and be published in a private or public AI store. The age of scalable AI has finally started. We explore scalable AI cases in agritech, healthcare, manufacturing and smart cities.

Speakers
avatar for Jonathan Berte

Jonathan Berte

CEO, Robovision
Jonathan Berte, CEO at Robovision, is a civil physics engineer (University of Ghent 2002). During his studies Jonathan specialised in applied neurology at the Institute of Neuroinformatics (INI). He is a pioneer in Artificial intelligence and saw the potential cultural impact early... Read More →



Thursday October 1, 2020 9:00am - 9:30am PDT
Virtual 4
  Case Studies
  • Slides Included Yes

9:00am PDT

Ray Internals: A Peek at `ray.get` - Stephanie Wang, Anyscale
The ability to reference remote memory is a cornerstone of the Ray API. In this talk, I'll give an overview of the lifetime of a Ray object. Along the way, I'll show what exactly happens during a `ray.get` call. I'll also explain how an object gets allocated, tracked, and deallocated. Finally, I'll give a quick overview of some of our current and future projects towards object persistence and distributed memory management.

Speakers
avatar for Stephanie Wang

Stephanie Wang

Software Engineer, Anyscale
Stephanie is a PhD student in computer systems at UC Berkeley, a software engineer at Anyscale, and a Ray contributor. She's interested in building robust and usable distributed systems for the average developer and likes provable guarantees.


Thursday October 1, 2020 9:00am - 9:30am PDT
Virtual 1

9:00am PDT

AlphaDow: Reinforcement Learning for Industrial Production Scheduling - Adam Kelloway, Dow Chemical
Adam has deployed reinforcement learning trained agents that optimize a production planning problem at Dow, Inc and deliver measurable business impact. Inspired by the successes of AlphaGO, this ambitious project is named AlphaDow. The production scheduling problem at Dow is non-trivial. Dow’s continuous manufacturing plants transition through a product sequence which must minimize the costs of transitions and maximize product availability for customers all whilst minimizing inventory on hand. Human schedulers use experience and heuristics to select the best sequence today. The trained RL agents augment this decision-making process. Adam has used Ray extensively to distribute his training among Azure based compute clusters. This significantly decreased the time needed to train an RL agent capable of producing optimal production schedules. Adam will present several of the challenges faced when training and deploying RL agents for industrial production scheduling and how Ray has helped him to overcome those challenges by facilitating the scale required to solve these complex problems and achieve real-world business impact.

Speakers
avatar for Adam Kelloway

Adam Kelloway

Innovation Manager, Dow Chemical
Adam Kelloway works as an Innovation Manager for Dow's Digital Fulfillment Center. (DFC) The DFC is leading the development and deployment of AI/ML technologies within Dow's integrated supply chain. Adam has previously worked on optimized production scheduling using Mixed Integer... Read More →



Thursday October 1, 2020 9:00am - 9:30am PDT
Virtual 2
  Reinforcement Learning
  • Slides Included Yes

9:00am PDT

State Management in Distributed Online Learning with Ray - Edmon Begoli, Oak Ridge National Laboratory
Distributed online learning systems are a machine learning systems that learn in a real-time, over a continuously arriving data and in a distributed manner. The challenges of distributed online learning are many, and all are non trivial. In this talk, we'll focus on the challenge of consistent and scalable maintenance of the state of learning in a fully distributed manner, with a special emphasis on how we used Ray actors to facilitate propagation of learning parameters, conflict resolution, and transactional consistency of updates. We will discuss our technical approaches and challenges in the context of our work on real-time suicide prevention predictive algorithms, fraud detection, and infectious disease surveillance.

Speakers
avatar for Edmon Begoli

Edmon Begoli

Director - Scalable Protected Data Facilities (SPDF), National Center for Computational Sciences (NCCS), Oak Ridge National Laboratory
Edmon Begoli, PhD is a director of ORNL's organization for research on protected data (SPDF), where he is responsible for research and design of large-scale systems for resilient and reliable computing on protected data. He also serves as the Principal Investigator (PI) for the joint... Read More →



Thursday October 1, 2020 9:00am - 9:30am PDT
Virtual 3

9:35am PDT

Distributed Black-Box Model Explanation with Ray - Alexandru Coca, Seldon
Advances in computer architecture have led to state-of-the-art performance of machine learning models in fields such as text and image classification, disease detection, to name a few. Deploying these models remains challenging in finance or medicine, where a significant risk element is associated with the decision making process. To bolster adoption of such systems, the AI community has recently focused on developing explanation models, which aim to help AI-assisted systems users trust the algorithms by using human-interpretable concepts to explain the model output. The explanation model uses an expensive search procedure, so explaining at scale is challenging. This talk shows how Ray can be used to distribute explanations on a Kubernetes cluster and thus reduce the time needed to explain multiple instances. Ray can thus be used to explain a model's behaviour on a large dataset, allowing data scientists to draw insights into their models. Crucially, the explanations of the model behaviour across an entire dataset can be indicative of biases in the training data. Thus, by leveraging Ray data scientists can ensure the systems they develop are fair and transparent.

Speakers
avatar for Alexandru Coca

Alexandru Coca

Research Engineer, Seldon.io
Alex has graduated with a System Engineering MEng degree in 2015. Following a two-year stint at Jaguar Land Rover where he was responsible for the system integration of complex semi-autonomous driving functions as part of their Highly Automated Driving Team, he has decided to pursue... Read More →



Thursday October 1, 2020 9:35am - 10:05am PDT
Virtual 3
  Case Studies
  • Slides Included Yes

9:35am PDT

RLlib: Scalable RL for TensorFlow, PyTorch, and Beyond - Eric Liang, Anyscale
Reinforcement learning is emerging as a practical tool for optimizing complex, unpredictable environments that can be simulated. For example, game artificial intelligence, system control, robotics, supply chain management, and finance. RLlib is an open-source library for reinforcement learning that offers both high scalability and a unified API for a variety of applications. In this talk we give an overview of RLlib, its design, and upcoming features.

Speakers
avatar for Eric Liang

Eric Liang

Software Engineer, Anyscale
Eric Liang is a software engineer at Anyscale and a graduate student at U.C. Berkeley, where he does research on reinforcement learning.



Thursday October 1, 2020 9:35am - 10:05am PDT
Virtual 1
  Ray and Its Libraries
  • Slides Included Yes

9:35am PDT

Project Bonsai - How Microsoft is Using Ray and RLlib Inside a Machine Teaching Service for Autonomous Systems - Edilmo Palencia, Microsoft
Machine Learning and, in special, Reinforcement Learning could be hard. But we believe that is possible to make it accessible to subject matter experts in any industry, regardless of knowledge computer science or Machine Learning.

At Microsoft Project Bonsai we are working in Machine Teaching, a new complementary approach to Machine Learning that could be used by those without AI expertise, to build AI agents that powers intelligent control systems through a unique combination of advance machine learning techniques, simulation, and deployment capabilities.

In this presentation we are going to talk about how we use RLlib and Ray to achieve our vision. And we will perform a quick demo of the platform using a special simulator created specially for learning about Bonsai and Machine Teaching.

Speakers
avatar for Edilmo Palencia

Edilmo Palencia

Principal AI Engineer, Microsoft



Thursday October 1, 2020 9:35am - 10:05am PDT
Virtual 2
  Reinforcement Learning
  • Slides Included Yes

9:35am PDT

Cloudstate—Towards Stateful Serverless - Jonas Bonér & James Roper, Lightbend
The Serverless experience is revolutionary and will grow to dominate the future of Cloud. Function-as-a-Service (FaaS) however—with its ephemeral, stateless, and short-lived functions—is only the first step. FaaS is great for processing-intensive, parallelizable workloads, moving data from A to B providing enrichment and transformation along the way. But it is quite limited and constrained in what use-cases it addresses well, which makes it very hard/inefficient to implement general-purpose application development and distributed systems protocols.

What’s needed is a next-generation Serverless platform and programming model for general-purpose application development in the new world of real-time data and event-driven systems. What is missing is ways to manage distributed state in a scalable and available fashion, support for long-lived virtual stateful services, ways to physically co-locate data and processing, and options for choosing the right data consistency model for the job.

This talk will discuss the challenges, requirements, and introduce you to our proposed solution: Cloudstate—an Open Source project building the next generation Stateful Serverless and leveraging state models such as Event Sourcing, CQRS, and CRDTs, running on Akka, gRPC, Kubernetes, and GraalVM, in a polyglot fashion with support for Go, JavaScript, Java, Swift, Scala, Python, Kotlin, and more.

Speakers
avatar for James Roper

James Roper

Cloud Architect, Lightbend
James is a long time open source contributor and Reactive systems expert. He is the creator of Cloudstate, the framework that brings distributed state management to the serverless world. He also created the Lagom Reactive microservices framework and is a core contributor to Play... Read More →
avatar for Jonas Bonér

Jonas Bonér

CTO, Lightbend
Jonas Bonér is founder and CTO of Lightbend, creator of the Akka project, initiator and co-author of the Reactive Manifesto, Chair of the Reactive Foundation, and a Java Champion. Learn more at: http://jonasboner.com... Read More →



Thursday October 1, 2020 9:35am - 10:05am PDT
Virtual 4

10:10am PDT

Dask-on-Ray: Using Dask and Ray to Analyze Petabytes of Remote Sensing Data - Clark Zinzow, Descartes Labs
Dask is a parallel computing library for Python, providing parallel ndarray and dataframe abstractions that mimic the interfaces of NumPy and Pandas. Beneath the interfaces of these abstractions, Dask automatically partitions data and composes a task graph representing NumPy and Pandas operations over these data chunks. This task graph is then executed by a Dask scheduler, parallelizing the execution of tasks across cores and/or machines, yielding both task and data parallelism.

This talk is about a Dask-on-Ray scheduler that was recently added to Ray. By implementing a Dask scheduler that farms Dask tasks out to a Ray cluster, we can run the entirety of the Dask ecosystem on top of Ray. By leveraging Ray’s innovations around decentralized peer-to-peer resource-aware locality-aware local-first scheduling, scheduling decision caching for fast worker-to-worker RPCs, zero-copy intra-node data sharing, task, worker, and GCS fault-tolerance, and scheduling throughput being unaffected by cluster state inspection, we obtain a fast, scalable, fault-tolerant Dask backend with great ops tooling at no performance cost, capable of executing any Dask graph.

Finally, we’ll talk about how Descartes Labs is leveraging Ray and this Dask-on-Ray integration to provide a geospatial data analysis product, Workflows. Using a highly scalable Ray-based backend, Workflows provides both low-latency interactive data analysis and massively scalable batch computation, distributing work across tens of thousands of cores and allowing users to process petabytes of remotely sensed data with ease

Speakers
avatar for Clark Zinzow

Clark Zinzow

Software Engineer, Descartes Labs
Clark Zinzow is a software engineer at Descartes Labs, where he's building a platform for geospatial data analysis that puts tens of thousands of cores and petabytes of remote sensing data at the user's fingertips. He loves working at the boundaries of service, data, and compute scale... Read More →



Thursday October 1, 2020 10:10am - 10:40am PDT
Virtual 4
  Ray and Its Libraries
  • Slides Included Yes

10:10am PDT

What's New with Ray Libraries: Tune - Richard Liaw, Anyscale
Hyperparameter tuning demands a lot of compute resources, so optimization is key. Tune leverages Ray for efficient hyperparameter tuning. This talk discusses the features of Tune, especially recent improvements, and provides a roadmap for future improvements.

Speakers
avatar for Richard Liaw

Richard Liaw

Software Engineer, Anyscale
Richard Liaw is a software engineer at Anyscale and a graduate student at U.C. Berkeley.


Thursday October 1, 2020 10:10am - 10:40am PDT
Virtual 1

10:10am PDT

Data Scientists are More Valuable than Hardware: The Design of Modin - Devin Petersohn, Intel
As tools trend toward scaling to larger and larger data, their requirements are growing as well, which presents a non-trivial human cost. To bring the focus of data science tools back to the data scientist, we built Modin, a platform that scales data scientist capabilities without requiring them to learn about distributed computing concepts. In this talk, we discuss the design of Modin and how Ray enables our modular design to scale. We also discuss our experience with enabling the data scientist to connect to multiple Ray clusters from the same notebook.

Speakers
avatar for Devin Petersohn

Devin Petersohn

Machine Learning Engineer, Intel
Devin Petersohn is a 5th year Computer Science PhD student at the UC Berkeley RISELab and Machine Learning Engineer at Intel. The early focus of his PhD work was in scaling genomic workloads to enable large scale DNA analysis. In recent years, Devin has focused on making scalable... Read More →


Thursday October 1, 2020 10:10am - 10:40am PDT
Virtual 3

10:10am PDT

Utilizing Reinforcement Learning Techniques for Supply Chain with Amazon SageMaker - Anna Luo, AWS
Reinforcement learning jobs are computationally expensive requiring multiple CPU and GPU instances. Researchers and practitioners need to manually set up the instances and manage utilization, especially in production. This process can be time consuming and adds additional operational overhead. Amazon SageMaker RL allows customers to focus on their RL research without the need to manage servers. It provides pre-built containers that supports RLlib with both TensorFlow and PyTorch framework. We will present how you can bring in your RL problem and use Amazon SageMaker RL to perform the training. You can easily manage and reproduce your experiments with either single or multiple instances. For production use, we will showcase how you can deploy the trained model with a single click and monitor the endpoint status.

Speakers
avatar for Anna Luo

Anna Luo

Applied Scientist, Amazon Web Services
Anna Luo is an Applied Scientist in the AWS. She works on utilizing reinforcement learning techniques for different domains including supply chain and recommender system, in which Ray is extensively used. She received her Ph.D. in Statistics from University of California, Santa Barbara... Read More →


Thursday October 1, 2020 10:10am - 10:40am PDT
Virtual 2

10:40am PDT

Stretching Break: Lower Back & Spine - Maryam Sharifzadeh, Office Yoga
Join us for 15 minutes of stretching.  No equipment required, do it right at your desk!

Thursday October 1, 2020 10:40am - 10:55am PDT
Virtual

10:40am PDT

Break
Thursday October 1, 2020 10:40am - 11:10am PDT
Virtual

10:40am PDT

Sponsored Office Hours: Anyscale
Anyscale was founded by the creators of Ray, an open source project from the UC Berkeley RISELab. Anyscale enables developers of all skill levels to easily build applications that run at any scale, from a laptop to a data center.

Thursday October 1, 2020 10:40am - 11:10am PDT

11:10am PDT

Human-First AI: Because AI Needs The Human Eye - Christopher Nguyen, Arimo
Today's Industrial-AI faces 3 key challenges: “Can't Start”, “Can't Profit”, and “Can't Deploy”. Christopher will share how his team has successfully overcome these challenges using “Human-First AI”, an open-source, collaborative data-science framework that leverages Ray at scale.

Speakers
avatar for Christopher Nguyen

Christopher Nguyen

President & CEO, Arimo
Christopher Nguyen is President & CEO of Arimo (acquired by Panasonic), and has led his team to successfully commercialize real-world, industrial-AI (manufacturing, avionics, robotics, energy, and automotive). He has served as engineering director of Google Apps and co-founded three... Read More →


Thursday October 1, 2020 11:10am - 11:20am PDT
Virtual 4

11:10am PDT

Project Zouwu: Scalable AutoML for Telco Time Series Analysis using Ray and Analytics Zoo - Ding Ding, Intel
Time series analysis plays a crucial rule in the telecom applications, such as network quality analysis, network capacity forecast, smart power management, etc. There’s a recent trend to apply machine learning methods (especially neural networks) to such problems, and they are reported to perform better in many cases than traditional methods such as autoregression and exponential smoothing.

However, building the machine learning applications for time series forecasting can be a laborious and knowledge-intensive process. In this talk, we present Project Zouwu, which provides Automated Machine Learning (AutoML) to time series analysis for Telco application. It is built on top of Ray (https://github.com/ray-project/ray) and Analytics Zoo (https://github.com/intel-analytics/analytics-zoo), so as to automate the process of feature generation and selection, model selection and hyper-parameter tuning in a distributed fashion. We will also share some real-world experience and “war stories” of earlier users.

Speakers
avatar for Ding Ding

Ding Ding

Machine Learning Engineer, Intel
Ding Ding is a machine learning engineer in Intel’s ML solution platform team, where she works on developing distributed machine learning and deep learning algorithms. She is an active contributor to the BigDL and Analytics Zoo projects.



Thursday October 1, 2020 11:10am - 11:40am PDT
Virtual 3
  Case Studies
  • Slides Included Yes

11:10am PDT

Introducing Ray Serve: Scalable and Programmable ML Serving Framework - Simon Mo, Anyscale
After data scientists train a machine learning (ML) model, the model needs to be served for interactive scoring or batch predictions. The go-to solution is often to wrap the model inside a Flask microservice. But when is that not enough? In this talk, I will discuss the short-comings of the Flask-only solution and then discuss the more common alternative, the “tensor prediction service” approach used by TFServing, SageMaker, and others. I will then introduce an easy-to-use, scalable ML serving system “Ray Serve” that overcomes the deficiencies of the two approaches. I will highlight the architectural innovations in Ray Serve.

Speakers
avatar for Simon Mo

Simon Mo

Software Engineer, Anyscale
Simon Mo is a software engineer at Anyscale. Before Anyscale, he was a student at UC Berkeley participating in research at the RISELab. He focuses on studying and building systems for machine learning, in particular, how to make ML model serving systems more efficient, ergonomic... Read More →



Thursday October 1, 2020 11:10am - 11:40am PDT
Virtual 1
  Ray and Its Libraries
  • Slides Included Yes

11:10am PDT

Connecting Reinforcement Learning to Simulation Software - Max Pumperla, Pathmind
Simulation software like AnyLogic is frequently used across many industries to model complex workflows of real-world applications. You can use these models to solve actual problems, and many of the questions that customers have can be framed as reinforcement learning tasks. However, by design, most simulation software providers do not have an integration with tools like Ray RLlib to apply reinforcement learning to their models directly.

We show how Pathmind leverages the best of both worlds by connecting existing simulation models to modern reinforcement learning tools. In particular, we discuss the special challenges we face and how we use Ray to address them by walking you through a concrete example.

Speakers
avatar for Max Pumperla

Max Pumperla

Deep Learning Engineer, Pathmind
Max is a deep learning engineer and prolific open source contributor. He’s maintainer of Hyperopt, DL4J core developer, Keras contributor and author of several Python libraries. He’s the author of “Deep Learning and the Game of Go” and Coursera instructor for “Applied AI... Read More →



Thursday October 1, 2020 11:10am - 11:40am PDT
Virtual 2
  Reinforcement Learning
  • Slides Included Yes

11:45am PDT

Turbocharging State-of-the-art Natural Language Processing on Ray - David Talby, John Snow Labs
This session introduces the Python nlu library, which provides the full power of Spark NLP as simple Python one-liners that directly read and write data frames. We will walk through some of the 250+ pre-trained NLP models & pipelines that come with the library. We'll then describe how the nlu library integrates Ray and Spark NLP to enable you to get the performance, scale, and accuracy benefits of both without having to learn new API's or implementation details.

Speakers
avatar for David Talby

David Talby

Chief Technology Officer, John Snow Labs
David Talby is a chief technology officer at John Snow Labs, the creators of Spark NLP: a production-grade, fast & trainable implementation of the latest research in natural language processing. David specializes in building & operating AI systems in healthcare and life science, and... Read More →



Thursday October 1, 2020 11:45am - 12:15pm PDT
Virtual 3

11:45am PDT

Adopting Ray @ LinkedIn - Jonathan Hung & Nitin Pasumarthy, LinkedIn
In this talk, we will present our effort to run Ray on YARN, and the integration of Ray on LinkedIn’s open-sourced offline infrastructure: Azkaban (workflow scheduling service) and TonY (Tensorflow on Yarn). We will provide a demo of running a Ray job end-to-end, discuss the architectural decisions and talk about our cooperation with the Ray team on this effort.
In the later half of the talk, we will share how we used Ray Tune on Kubernetes in a real world use case. Tune helped us identify promising model configurations, using state of the art Bayesian optimization algorithms like TPE & PBT, with minimal supervision. The data pipeline is also optimized using Tune to extract the maximum throughput and we were able to train 2x faster by reducing the GPU idle times.

Speakers
JH

Jonathan Hung

Senior Software Engineer, LinkedIn
Jonathan Hung is a senior software engineer on the Hadoop development team at LinkedIn.



Thursday October 1, 2020 11:45am - 12:15pm PDT
Virtual 2
  Ray in the Enterprise
  • Slides Included Yes

11:45am PDT

Debugging and Observability for Distributed Ray Applications - SangBin Cho, Anyscale
Developing distributed applications is difficult, and Ray tries to make it easier. However, it’s a well-known fact that most of our time is spent debugging, not actually writing the initial code. This talk will describe the vision for making debugging applications during development and monitoring them in production simple when using Ray. First, we’ll describe the recent improvements that we’ve made to metrics monitoring and the Ray dashboard. Then, we’ll present a few possible future directions including dashboard improvements, structured events, and distributed tracing. This will include a short demo of how to debug an application using these tools.

Speakers
avatar for SangBin Cho

SangBin Cho

Software Engineer, Anyscale
SangBin Cho is a software engineer at Anyscale


Thursday October 1, 2020 11:45am - 12:15pm PDT
Virtual 1

12:20pm PDT

Ray + Arize - Why you Need Model Observability - Aparna Dhinakaran, Arize AI
As more and more machine learning models are deployed into production, it is imperative we have better observability tools to monitor, troubleshoot, and explain their decisions. In this talk, Aparna Dhinakaran, Co-Founder, CPO of Arize AI (Berkeley-based startup focused on ML Observability), will discuss the state of the commonly seen ML Production Workflow and its challenges. She will focus on the lack of model observability, its impacts, and how Arize AI can help.  
This talk will highlight common challenges seen in models deployed in production, including model drift, data quality issues, distribution changes, outliers, and bias. The talk will also cover best practices to address these challenges and where observability and explainability can help identify model issues before they impact the business. Aparna will be sharing a demo of how the Arize AI platform can help companies validate their models performance, provide real-time performance monitoring and alerts, and automate troubleshooting of slices of model performance with explainability. The talk will cover best practices in ML Observability and how companies can build more transparency and trust around their models.

Speakers
avatar for Aparna Dhinakaran

Aparna Dhinakaran

Chief Product Officer, Arize AI
Aparna Dhinakaran is Chief Product Officer at Arize AI, a startup focused on ML Observability. She was previously an ML engineer at Uber, Apple, and Tubemogul (acquired by Adobe). During her time at Uber, she built a number of core ML Infrastructure platforms including Michaelangelo... Read More →


Thursday October 1, 2020 12:20pm - 12:50pm PDT
Virtual 3

12:20pm PDT

Advanced Ray RLlib: New Features and New Possibilities - Sven Mika, Anyscale
Deep Reinforcement Learning (DRL) has become the technology of choice when trying to autonomously learn optimal policies for agents interacting with each other in arbitrarily complex environments. RLlib - Ray's open source RL library - addresses the problems of finding well performing algorithms for a given environment, configuring and customizing them, as well as running these algos in large-scale, distributed workloads to yield good decision making policies. This talk focuses on features and algorithms recently added to RLlib and how you can use these to solve novel environment- and problem types. Topics include in particular: Model-based RL, a new exploration API for RLlib, PyTorch support, curiosity and intrinsic motivation, and contextual bandits.

Speakers
avatar for Sven Mika

Sven Mika

Machine Learning Engineer, Anyscale
Sven is a machine learning engineer at Anyscale Inc. and lead developer of Ray RLlib, one of the most popular reinforcement learning open-source libraries. He is responsible for implementing recently published, promising algorithms, designing user-friendly APIs, asserting high levels... Read More →


Thursday October 1, 2020 12:20pm - 12:50pm PDT
Virtual 1

12:20pm PDT

Information Gain Regulation in Reinforcement Learning with the Digital Twins' Level of Realism In RlLib - Géza Szabó, Ericsson Research & József Pető, Budapest University of Technology and Economics
Digital Twin (DT) is widely used in various industrial sectors to optimize the operations and maintenance of physical assets, system and manufacturing processes. Our goal is to introduce an architecture in which the radio access control happens automatically to minimize the utilized radio resources while still maximizing the production KPIs of the robot cell. To achieve this, we apply Reinforcement Learning (RL) in a simulated environment to explore the environment fast, while the DT ensures that the learned policy can be applied on the real world environment as well. We show that the application of Ultra Reliable Low Latency Communication (URLLC) connection can be reduced to approx. 30% of the total radio time while achieving real-world accurate robot control. We discuss the implementation of the concept within RLlib.

Speakers
avatar for Géza Szabó

Géza Szabó

Senior Researcher, Ericsson Research
Géza Szabó joined Ericsson Research as an undergraduate student in 2005. He wrote his MSc thesis on Ericsson topic in the Budapest University of Technology and Economics, about comparing various application traffic classification methods. He pursued further this topic in his PhD... Read More →
avatar for József Pető

József Pető

PhD Student, Budapest University of Technology and Economics
József Pető received his MSc. degree in Computer Engineering from the Budapest University of Technology and Economics in 2018.He is currently working toward his Ph.D. degree. His current areas of interest and research include cloud robotics, digital twin, robot simulation, machine... Read More →


Thursday October 1, 2020 12:20pm - 12:50pm PDT
Virtual 2

12:50pm PDT

Afternoon Break
Thursday October 1, 2020 12:50pm - 4:00pm PDT
Virtual

1:00pm PDT

BoF: Reinforcement Learning
Interested in RL or already using it? Discuss how you're using reinforcement learning, what tools you're using, and what challenges you face.

Join in Zoom. 

Thursday October 1, 2020 1:00pm - 1:45pm PDT

1:45pm PDT

BoF: Ray
Interact with fellow Ray users and contributors! Discuss how you're building distributed applications with Ray and Ray’s ecosystem of libraries.

Join in Zoom.

Thursday October 1, 2020 1:45pm - 2:30pm PDT

2:30pm PDT

BoF: Python Data
Python has emerged as the dominant programming language for machine learning and data science. Discuss how you're scaling Python for data processing, visualization, machine learning, and more.

Join in Zoom.

Thursday October 1, 2020 2:30pm - 3:15pm PDT

3:15pm PDT

BoF: Natural Learning Processing (NLP)
NLP and NLU are among the most exciting areas in AI. Discuss how you're using natural language technologies, what tools you are using, and the challenges you face.

Join in Zoom.

Thursday October 1, 2020 3:15pm - 4:00pm PDT

4:00pm PDT

Keynote: A Secure Collaborative Learning Platform - Raluca Ada Popa, Computer Security Professor, UC Berkeley
Multiple organizations often wish to aggregate their sensitive data and learn from it, but they cannot do so because they cannot share their data. For example, banks wish to train models  jointly over their aggregate transaction data to detect money launderers because criminals hide their traces across different banks. To address such problems, my students and I developed MC^2, a framework for secure collaborative computation. My talk will overview our MC^2 platform, from the technical approach to results and adoption.

Speakers
avatar for Raluca Ada Popa

Raluca Ada Popa

Computer Security Professor, UC Berkeley


Thursday October 1, 2020 4:00pm - 4:15pm PDT
Virtual

4:15pm PDT

Keynote: Every Five Decades: A New Field of Engineering - Michael I. Jordan, Distinguished Professor, University of California at Berkeley
Machine learning can be viewed as heralding the emergence of a new field of engineering, in a manner similar to the emergence of chemical engineering from the chemistry, thermodynamics, and fluid mechanics, or to the emergence of electrical engineering from electromagnetism. In both cases, new mathematical and conceptual tools were needed to support the ambitions of the new engineering field, allowing real-world systems to be built at scale.

In the case of machine learning, the subject matter is the blending of data, inference, and computing that we see in emerging systems and business models that we find in areas such as commerce, transportation, entertainment, and health care. The underlying scientific foundations are provided by the computational and inferential disciplines. But new mathematical and conceptual tools are needed to allow this field to emerge, and to provide support for building large-scale systems that are understandable, robust, transparent, safe, and useful. Most notably, while earlier engineering fields focused on physical materials, the new field focuses on data, decisions, and human context.

Speakers
avatar for Michael I. Jordan

Michael I. Jordan

Distinguished Professor, University of California, Berkeley.
Michael I. Jordan is the Pehong Chen Distinguished Professor in the Department of Electrical Engineering and Computer Science and the Department of Statistics at the University of California, Berkeley.His research interests bridge the computational, statistical, cognitive and biological... Read More →


Thursday October 1, 2020 4:15pm - 4:30pm PDT
Virtual

4:30pm PDT

Keynote: AlphaStar: How Google Infrastructure Made it Possible - Oriol Vinyals, Principal Scientist, Google DeepMind
Games have been used for decades as an important way to test and evaluate the performance of artificial intelligence systems. As capabilities have increased, the research community has sought games with increasing complexity that capture different elements of intelligence required to solve scientific and real-world problems. In recent years, StarCraft, considered to be one of the most challenging Real-Time Strategy (RTS) games and one of the longest-played esports of all time, has emerged by consensus as a “grand challenge” for AI research.

In this talk, I will introduce our StarCraft II program AlphaStar, the first Artificial Intelligence to reach Grandmaster status without any game restrictions. The focus of the talk will be on the scale and infrastructure that made it possible, highlighting not only the research but also engineering challenges behind the project.

Speakers
avatar for Oriol Vinyals

Oriol Vinyals

Principal Scientist, Google DeepMind
Oriol Vinyals is a Principal Scientist at Google DeepMind, and a team lead of the Deep Learning group. His work focuses on Deep Learning and Artificial Intelligence. Prior to joining DeepMind, Oriol was part of the Google Brain team. He holds a Ph.D. in EECS from the University of... Read More →


Thursday October 1, 2020 4:30pm - 4:45pm PDT
Virtual

4:50pm PDT

Distributed Deep Learning with Horovod on Ray - Travis Addair, Uber
Horovod is an open source framework created to make distributed training of deep neural networks fast and easy for TensorFlow, PyTorch, and MXNet models.  Horovod's API makes it easy to take an existing training script and scale it run on hundreds of GPUs, but provisioning a Horovod job with hundreds of GPUs can often be a challenge for users who lack access to HPC systems preconfigured with tools like MPI.  The newly introduced Elastic Horovod API introduces fault tolerance and auto-scaling capabilities, but requires further infrastructure scaffolding to configure.  In this talk, you will learn how Horovod on Ray can be used to easily provision large distributed Horovod jobs and take advantage of Ray's auto-scaling and fault tolerance with Elastic Horovod out of the box.  With Ray Tune integration, Horovod can further be used to accelerate your time-constrained hyperparameter search jobs. Finally, we'll show you how Ray and Horovod are helping to define the future of machine learning workflows at scale.

Speakers
avatar for Travis Addair

Travis Addair

Senior Software Engineer II, Uber Technologies
Travis Addair is a software engineer at Uber working on the Michelangelo machine learning platform. He leads the Horovod project and chairs its Technical Steering Committee within the Linux Foundation.  In the past, he’s worked on scaling machine learning systems at Google and... Read More →



Thursday October 1, 2020 4:50pm - 5:20pm PDT
Virtual 4
  Case Studies
  • Slides Included Yes

4:50pm PDT

Building Complex Data Analytics Pipelines with Ray - Qingqing Mao, Dascena
Building scalable data analytics pipelines is challenging, especially when different subtasks may have different computational requirements and interdependencies. It becomes more challenging when you need to serve enterprise customers who have strict data security and privacy policies and require on-premise deployment. The scaling requirement and computational capacity often vary widely from site to site.

We have been using Ray to build natural language processing pipelines and healthcare analysis pipelines. The highly efficient serialization using a shared-memory object store is a perfect fit for handling our data-intensive jobs. Ray helps us narrow the gap between data science and engineering, and it enables our data scientists to write high-performance and cost-efficient data analytics pipelines that can scale. 

Speakers
avatar for Qingqing Mao

Qingqing Mao

Head of Engineering and Data Science, Dascena
Qingqing Mao is the Head of Engineering and Data Science at Dascena, where he leads the development of compliant and scalable clinical data pipelines and the research on applying machine learning techniques in healthcare and medicine. Previously, he worked as a senior staff data scientist... Read More →



Thursday October 1, 2020 4:50pm - 5:20pm PDT
Virtual 2

4:50pm PDT

Leveraging Ray to Enhance Machine Learning Models in Finance - James Dunworth-Crompton, Goldman Sachs
Many of the issues faced when applying ML techniques to problems in finance greatly benefit from a robust framework for distributed applications like Ray. In this talk we will go over some of the interesting challenges that arise, from model selection and backtesting to gaining user confidence, and how Ray may be used overcome them.

Speakers
avatar for James Dunworth-Crompton

James Dunworth-Crompton

Associate, Goldman Sachs
James Dunworth-Crompton is an associate on the Core Machine Learning team at Goldman Sachs, working on bringing business value to the firm by apply machine learning techniques to a variety of unique problems across the firm's many businesses. Previously to working at GS James was... Read More →



Thursday October 1, 2020 4:50pm - 5:20pm PDT
Virtual 3
  Ray in the Enterprise
  • Slides Included Yes

4:50pm PDT

Using Ray On Large-scale Applications at Ant Group - Jiaying Zhou, Ant Group
As a FinTech company, Ant Group emphasizes the core value of digital payments platform as it underpins e-commerce and financial services such as loans and insurance. Behind the digital payments platform lies the big data system, which serves as a pillar for effective data analysis and decision-making. Therefore, Ant Group particularly concerns about the latency and reliability of distributed systems. Last year Ant Group successfully applied Ray in financial service, marketing-promotion, risk control and other scenarios in Double 11 Shopping Festival 2019. In this talk, Jiaying will share their experiences in performance assurance and stability control in this process.

Speakers
avatar for Jiaying Zhou

Jiaying Zhou

Senior Staff Engineer, Ant Group
Jiaying Zhou is a senior staff engineer at Ant Group focused on big data and computing. Since joining in 2011, he is responsible for several large-scale distributed production systems including real-time data platform, feature center, data eventing platform, and online computing system... Read More →



Thursday October 1, 2020 4:50pm - 5:20pm PDT
Virtual 1
  Ray in the Enterprise
  • Slides Included Yes

5:25pm PDT

ML Engineering Kubeflow - Pavel Dournov, Google Cloud
Kubeflow offers a Kubernetes native ML Platform stack with a wide spectrum of capabilities, and deployable anywhere Kubernetes runs. In this talk we will give an overview of the Kubeflow project and architecture, and will do a deeper dive into the ML Engineering practices for creating robust ML workflows and their implementation with Kubeflow Pipelines and Tensorflow Extended, leveraging the pipelines DSL and ML Metadata and automatic data lineage tracking.

Speakers
avatar for Pavel Dournov

Pavel Dournov

Engineering Manager, Google Cloud AI Platform
Pavel is an engineering manager in Google Cloud AI Platform, working on AI infrastructure services specialized in deploying and scaling AI workloads on Google Cloud. Pavel supports the teams building Kubeflow Pipelines and TFX on Cloud. Prior to Google Pavel helped build Azure Machine... Read More →



Thursday October 1, 2020 5:25pm - 5:55pm PDT
Virtual 2
  Case Studies
  • Slides Included Yes

5:25pm PDT

Java API and Cross Language Programming on Ray - Hao Chen, Ant Group
Ray was originally created as a framework for building distributed Python applications, mainly targeting at the AI area. At Ant Group, besides using Python, we also developed the Java API on top of Ray and use it to build large-scale big-data systems. In addition, we developed the cross-language invocation functionality, which allows you to seamlessly build a system that contains components written in different languages.

In this talk, we’ll introduce Ray’s Java API and show you how to write multi-language distributed applications with Ray.

Speakers
avatar for Hao Chen

Hao Chen

Staff Engineer, Ant Group
Hao Chen is a staff engineer at Ant Group and a Ray committer. At Ant, he leads the development of Ant’s internal fork of Ray, including many key features and architecture improvements that have been contributed back to the open source community, such as Java worker, actor fault... Read More →



Thursday October 1, 2020 5:25pm - 5:55pm PDT
Virtual 1

5:25pm PDT

Distributed Reinforcement Learning for Robotic Assembly - Rodger Luo, Autodesk Research
Robotic systems for automated assembly have been widely used in manufacturing, where the environment can be carefully and precisely controlled, but they are still in infancy in architectural construction. A main reason is that current robotic systems are not adaptive to the diversity of the real world, especially in unstructured settings. RL-based robotic systems are a promising direction given their adaptability to uncertainties.

This talk will introduce a group of projects from Autodesk Research to improve RL-based robotic systems for assembly tasks. We focus on robotic assembly tasks that involve contact forces, because such tasks are widespread in industrial applications and yet challenging for robots to do. All the algorithm developments are built on top of Ray’s framework.

Speakers
avatar for Rodger Luo

Rodger Luo

Principal AI Research Scientist, Autodesk Research
Rodger (Jieliang) Luo is a Principal AI Research Scientist at the Autodesk AI Lab in San Francisco. He received his Ph.D. from UCSB in 2020, where he explored the intersection of machine learning, robotics, and creativity. His current research focus is understanding how to learn complex... Read More →



Thursday October 1, 2020 5:25pm - 5:55pm PDT
Virtual 3
  Reinforcement Learning
  • Slides Included Yes
 
  • Timezone
  • Filter By Date Ray Summit 2020 Sep 29-Oct 1, 2020
  • Filter By Venue Virtual
  • Filter By Type
  • Anyscale Academy
  • BoF
  • Break
  • Case Studies
  • Keynote Sessions
  • Natural Language Processing
  • Ray and Its Libraries
  • Ray in the Enterprise
  • Reinforcement Learning
  • Research Meets Industry
  • Sponsored Office Hours
  • Slides Included