We4AI Lab — a personal lab notebook
I build AI agents and share what I learn.
Small, focused agents built as personal learning projects — published with the code, the costs, and the failures. No hype, no services, nothing for sale.
Featured work
Featured work
View all Project — Agent #1 active
We4AI Lab: Agent #1 — MCP Tool-Use Agent
Project geekyrahul.in — this website Method
Video code — public repos
costs — measured, dated
failures — written down
costs — measured, dated
failures — written down
About Who is GeekyRahul? Latest writing A Practical Guide to Migrating LLMs in Production RAG Systems Latest writing
Latest writing
View all A Practical Guide to Migrating LLMs in Production RAG Systems As the GPT-4 series sunsets, migrating RAG systems to GPT-5 isn't a 1-line swap. Here is the engineering playbook to handle behavioral drift and refusals. Why I'm building AI agents in public Why every agent I build gets a public page with its code, costs, and failures — and what that forces me to do differently. Never Re-Embed Twice: How an S3 Cache Saved My RAG Pipeline How decoupling vector generation from indexing with a cloud cache cut rebuilds to 2 seconds, eliminated API costs, and made DB migrations painless.
Contributions across platforms
Proof of work
GitHub profile317 contributions in the last year
GitHub commits, LeetCode submissions, and posts published here — merged per day, refreshed on every deploy.