Singapore Machine Learning Season 1 Episode 3

 16.01.2024 -  Singapore Machine Learning -  ~3 Minutes

When?

  • Tuesday, January 16, 2024 from 7:00 PM to 10:00 PM (Singapore Time)

Where?

  • In-person event: DataRobot, 11th Floor, 11-01, 5 Temasek Boulevard, Suntec Tower 5, Singapore.

The page of the event on Meetup: SGML S1E3

Thanks a lot to our sponsor DataRobot Singapore for hosting the event in its office! And also providing drinks and pizzas :)

Retrieval Augmented Generation for LLMs

Abstract: LLMs are becoming more and more heavily used in various industries and applications. To make them more powerful and reliable, researchers are now giving them a set of tools to use, such as a calculator, a web browser, or even a code interpreter! We will review recent research advancements in augmented generation during the past year to gain more insights.

Bio: Esther (Ruochen) Zhao, AI Researcher, Ph.D. Candidate at NTU. Her thesis focuses on Trustworthy and Interpretable NLP.

Slides

A Guide to AI Accelerators

Abstract: AI Accelerators are code-first, modular building blocks for model development and deployment using the DataRobot API. You can see them as powerful templates, available to the DataRobot community, that speed up the time to value. We will discuss how these are organized, show some popular examples and hopefully get you excited to contribute to the community.

Bio: João Gomes works for DataRobot as field CTO, where he does R&D work on new product features to accelerate value creation and advises businesses on how to successfully manage AI projects. Previously he was a principal investigator and lab head with the Institute for Infocomm Research (I2R) under the Agency for Science, Technology and Research (A*Star), Singapore. He holds a Ph.D. in computer science from the Technical University of Madrid.

Slides

LLM evaluation & LLM with Rust

Abstract:

Topic 1: Validating Large Language Models (LLMs)

Prompt Validation: • Techniques for effective prompt design. • Ensuring accurate and contextually relevant prompts. Qualitative Validation: • Assessing response accuracy and relevance. • Evaluating language model’s understanding and output coherence. Quantitative Validation: • Metrics for measuring performance. • Statistical approaches to validate LLM responses.

Topic 2: Leveraging Rust for LLM Inference

Why Rust? • Comparing Rust with Python in LLM contexts. • Advantages of Rust for high-performance computing. Inference Pipelines with Rust: • Introduction to the ‘candle’ library. • Building lightning-fast LLM inference pipelines. • Case studies or examples of Rust accelerating LLM processing.

Bio: Praveen (Govindaraj) is a seasoned Staff Data Scientist at Singtel. With 13 years of experience in leveraging data science to drive significant business improvements. Praveen’s expertise spans from customer engagement to anti-money laundering/countering financing of terrorism (AML/CFT), credit risk modeling, and data warehousing. Praveen has a proven record of reducing customer churn, streamlining processes, and enhancing business performance through data-driven strategies.

Slides

SGML S1E3 Crowd