Welcome🔗
I'm Wassim, ML Engineer at iAdvize in Paris. I build LLM & RAG systems at production scale evaluation, observability, retrieval, agentic architectures. This page is a curated entry point into my work and the resources I find most valuable.
Key Themes in My Work🔗
RAG Systems🔗
- Avoid Citation Hallucination -
Why incremental document IDs invite hallucination, and how to fix it
LLM Optimization & Evaluation🔗
Posts coming soon.
Engineering Philosophy🔗
- You Should Avoid AI Frameworks -
They abstract the wrong layer. Own your context, own your prompts.
Agentic Systems🔗
Posts coming soon.
Must-Read External Resources🔗
Resources I keep coming back to and recommend to anyone working on production AI.
Guides & Reference
- What We Learned from a Year of Building with LLMs -
Eugene Yan et al. Practical lessons from shipping LLM products - Applied LLMs -
Community-driven resource on building with LLMs - Anthropic's Prompt Engineering Guide -
Best practices for prompting - LMSYS Chatbot Arena -
Live LLM benchmarking by human preference
Papers
- Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks -
The original RAG paper (Lewis et al.) - Lost in the Middle -
How LLMs struggle with long contexts - Chain-of-Thought Prompting -
Wei et al., foundational prompting technique
Tools
- LangSmith -
LLM observability and evaluation - Ragas -
RAG evaluation framework - vLLM -
High-performance LLM serving
Projects & Competitions🔗
- Kaggle NFL Helmets Detection Silver Medal (Top 5%). Computer vision competition detecting helmets in NFL footage
- This blog Built with Material for MkDocs, deployed on GitHub Pages. Source code
Math🔗
My personal passion. I use math to build intuition about how things work from probability theory to optimization. See Math for Fun.