RAG, Fine-Tuning & What's Next
You can now train a custom AI system on your domain data, ship it as a feature, and speak fluently about what's coming next in AI.
3-hour live session · 90 min theory · 90 min hands-on · plus one exercise
The 3-hour live session
Theory grounds you. Hands-on lands it. Both happen in the same Saturday session.
Theory
Decision matrix day. RAG vs fine-tuning vs system prompt - when each one wins, and when each one wastes your budget. Then a 30-minute coda on what's coming next: A2A, NANDA, SLMs, GEO, llms.txt. The session that makes you feel early instead of late - and gives you the vocabulary for conversations with your CTO that nobody else on your team is having yet.
Hands-on
Build something that knows your domain better than ChatGPT. Real documents from your work, real retrieval, real answers. The portfolio piece you'll show in your next interview. By Saturday evening you've shipped Project 3 - and you've finished what 96 PMs before you have done.
Your exercise for the week
Build a RAG system on documents from your real work. Ask it ten questions your team gets stuck on. Bring the best three answers and the worst two.
The moment that lands
The podcast RAG system from Cohort 2: a popular product podcast (300 episodes) ingested into a vector DB. Ask 'how do you track source attribution in products?' - get a paragraph answer with episode citations. A Cohort 2 participant: "This is like Perplexity but for the one podcast I actually care about." A Cohort 3 participant replicated it independently over Christmas break.
What I'll keep saying
- “RAG is: search your docs, feed results to AI, get a smart answer. That is it.”
- “Agent card is a LinkedIn profile for agents. You publish what you can do. Other agents discover you and reach out.”
- “NANDA is DNS for agents. You type google.com, DNS fetches an IP. You query NANDA, it fetches an agent card.”
Want this week in your real life?
DM me on WhatsApp or book a 30-minute call. The cohort is 12 to 18 PMs and I personally vet every applicant.