Vince Lam

Vince Lam

Data Science & AI Engineering

H2O Singapore GenAI Day

1 minute read

Published:

I recently attended H2O.ai’s Singapore GenAI Day, a fantastic event showcasing their latest contributions to GenAI. Highlights included:

  1. H2O and GenAI Keynote: Insights and inspiration from Sri Satish Ambati, CEO and founder.

  2. Foundations of GenAI Ecosystems: Shivam Bansal comprehensively covered everything from transformers to LLM guardrails.

  3. LLM + RAG Lab: We all created our personal GenAI chatbot from our CVs using h2oGPTe with Llama2 and RAG, led by Chun Ming Lee.

  4. Building GenAI Apps: Genevieve Richards led the workshop on developing a web app using Wave Studio which integrated the previous model.

  5. LLM Fine-Tuning: Vishal Sharma, PHD’s walked through LLM Studio for model fine-tuning and seamless integration with HuggingFace.

  6. Data Preparation and Evaluation: Timothy CL Lam demonstrated LLM DataStudio and LLM Eval for efficient data preparation and model comparison.

H2O Founder and CEO delivering the GenAI keynote

H2O Founder and CEO, Sri, delivering the GenAI keynote

The standout tools for me were LLM Studio for practical fine-tuning and LLM Eval for model assessment, both open-source. The upcoming RAG evaluation release is something I’m looking forward to, enhancing QA systems for real-world production. Excited to see how H2O’s innovations continue to shape the GenAI ecosystem.

What Are Embeddings and Why They’re Incredibly Useful

1 minute read

Published:

If you’ve heard of the term “embeddings” before but don’t quite understand it, this post is for you. I’ll provide an intuitive explanation of embeddings and their significance.

Traditional Natural Language Processing (NLP) Tasks: Understanding What Large Language Models (LLMs) Can Do For You

1 minute read

Published:

Since Transformers were introduced in 2017, Natural Language Processing (NLP) capabilities have skyrocketed. Knowing traditional NLP tasks is useful to understand the capabilities of (Large) Language Models. Although LLMs may outperform in many areas, smaller fine-tuned models will always have a place due to their faster latency and lower costs.