Free live masterclass · 90 minutes
Shipping RAG That Doesn't Break in Production
A free live session for Python engineers who've built a RAG demo that works on their laptop — and now need it to work in front of users.
Sunday, October 11, 2026 · 4:00 PM IST
Live on Zoom · 90 minutes · Recorded
Doors open 15 min early — come check your audio
Save my seat
Replay sent to everyone who registers — even if you can't make it live.
What you'll walk away with
- Why pure vector search fails in production (and the hybrid pattern that fixes it)
- A chunking strategy that doesn't lose meaning and doesn't balloon your token bill
- How to evaluate a RAG pipeline without kidding yourself (retrieval vs generation)
- The real latency + cost math — measured, not guessed
- A debugging playbook for 'why did my RAG return garbage for this specific query'
Who this is for
You'll get the most out of this if:
- You've built a RAG demo and are now asked to put it in front of paying users
- You're a Python engineer who can call the OpenAI / Anthropic API and want to build beyond it
- You're tired of 'RAG in 10 lines of code' threads and want the actual production shape
Not for you if:
- You're brand new to LLMs
- You want a tutorial on calling an API
Agenda
0:00Live: a RAG pipeline returning wrong answers, diagnosed live
0:10Embeddings, chunking, retrieval, reranking — the production shape
0:50Evals that actually tell you if things got better
1:05Q&A
1:20What the 9-weekend cohort covers (optional to stay)
Your host
Arnab Pal
Engineer and founder. Has shipped production RAG and agent features on Anthropic + OpenAI with pgvector — for content platforms, customer support, and internal tooling. Runs Stackedtensor and arnabpal.me.
FAQ
Save your seat
90 minutes live · Recording sent to everyone who registers
Already know you want to go deeper? See the full 9-weekend cohort →