Bay Area AI + Ray Summit Meetup
Details
Our next in-person meetup is held on the last day of the Ray Summit 2023.
We focus on Ray use in AI, such as LLM, and deployment of ML workloads in various use cases. Weights & Biases will host and present, and planned talks include IBM Research and Open-Source Science. We'll also host lightning talks and networking.
Please register to attend here 👇:
👉 https://www.eventbrite.com/e/bay-area-ai-ray-summit-meetup-tickets-707923998737 👈
Speakers:
Anish Shah, MLE, Weights & Biases
Abstract: Are you looking for ways to improve the efficiency and success of your multi-agent reinforcement learning (MARL) experiments? Ray can automate MARL tuning by dynamically tracing and optimizing experiments across parameter spaces, while utilizing your computational resources via its tools such as AIR, Tune, and RLLib. Combined with Weights & Biases, MARL engineers get a comprehensive view of their experiments, centralizing all details and assets into one ML system of record. This enables them to save time on experiment iterations and quickly determine which tunings produce optimal performance outcomes, providing faster results with fewer hiccups. To demonstrate this, we will perform experiments on two scenarios: Autonomous Vehicle Driving and Drone Flying.
Bio Anish loves turning ML ideas into ML products. He began his career working with several Data Science teams at SAP, utilizing traditional Machine Learning, Deep Learning, and building recommendation systems. Now, he is at Weights & Biases, engaging with practitioners to create the right tools, lessons, and collateral to make Machine Learning accessible to everyone. With the art of programming and a little bit of magic, Anish crafts ML projects to help better serve others, turning "oh no's" into "a-ha's"!
Charles Fan, CEO, MemVerge
Abstract:
MemVerge will introduce its vision for accelerating AI/ML workflows by consolidating distributed memory object pools into a centralized shared memory object pool. The goal is to reduce overhead from memory copies, network communication, and I/O serialization/deserialization. Initial prototypes show a 700% acceleration. MemVerge has submitted a REP for creating a pluggable memory object layer interface into Ray.
Bio:
Charles Fan is co-founder and CEO of MemVerge. Prior to MemVerge, Charles was the CTO of Cheetah Mobile leading its global technology teams, and an SVP/GM at VMware, founding the storage business unit that developed the Virtual SAN product. Charles also worked at EMC and was the founder of the EMC China R&D Center. Charles joined EMC via the acquisition of Rainfinity, where he was a co-founder and CTO. Charles received his Ph.D. and M.S. in Electrical Engineering from the California Institute of Technology, and his B.E. in Electrical Engineering from the Cooper Union.
Ofer Mendelevitch
Abstract: Retrieval augmented generation (RAG; also known as Grounded Generation or GG) is a process for augmenting LLM knowledge with your data, powering a broad class of use-cases for LLMs like question-answering and chatbots with reduced hallucinations. In this talk I will explain how RAG/GG works, explain why and how it reduces hallucinations of LLMs, and share a reference architecture for RAG/GGl. In particular, I'll share
one scalable reference architecture where Ray / Anyscale are employed for both pre-processing of data before embedding and also for serving of LLMs (Aviary / Anscale Endpoints).
I will then share some best practices for getting your RAG implementation done correctly - such as proper chunking, using the right embedding model, hybrid search and low latency LLM serving - and how to address some of the challenges of moving from a toy example to enterprise deployment. Finally I'll show a demo of an LLM-powered app (powered by the Vectara GenAI platform) for asking questions about recent news.
Bio: Ofer Mendelevitch leads developer relations at Vectara. He has extensive hands-on experience in machine learning, data science and big data systems across multiple industries, and has focused on developing products using large language models since 2019. Prior to Vectara he built and led data science teams at Syntegra, Helix, Lendup, Hortonworks and Yahoo! Ofer holds a B.Sc. in computer science from Technion and M.Sc. in EE
Sponsors: IBM, Microsoft, Weights & Biases, AnyScale, Vectara
--REMINDER--
For more incredible talks and speakers, you won't want to miss SBTB'23 in Oakland happening on Nov 14-15! The central theme is "Code and Data in the Age of AI." Enjoy workshops, a fireside chat with OpenAI, & network with industry leaders. Early Bird are still on sale 🌟📊🤖