Real answers. Clear concepts. From "the demo works on my machine" to "it ships and holds up in production" LLMs, RAG, agents, MCP & evals.
** Hello students! **
Here an AI Engineer with 9+ years building real, working systems. I built the MCP layer for Mercado Libre's AI assistant (typed, idempotent tools over real backend APIs), and I run my own AI stack in production: a RAG + long-term memory server with eval gates, and an agent harness I extend at the source. I know this stuff because I live it daily not from a course.
🧠 But here's what really matters:
AI is easy to demo and hard to ship. Anyone can make a chatbot in a weekend; keeping it correct, cheap, and honest in production is the actual job and it can really feel overwhelming.
Whether you're:
A beginner who thinks embeddings are witchcraft,
In a hackathon with a demo that melts the moment real users arrive,
Or a pro whose agent ignores half its instructions and burns your token budget
I'm here to help.
** How I Can Help **
📍 New to AI engineering?
Understand where LLMs fit and where a plain function beats one.
Prompts as engineering, not vibes: structure, context, and why they work.
Build your first RAG that actually retrieves the right things.
Stop copy-pasting AI code you don't understand i'll help you actually get it.
🎓 Building your first AI feature?
Chunking, embeddings, hybrid retrieval — explained without the hype.
Evals from day one: a small golden set beats a thousand good intentions.
Token budgets, context windows, and cost — before your bill does.
💼 Already shipping AI?
Design MCP tools the way production demands them: typed, idempotent, observable.
Agents that don't loop, don't hallucinate tools, and degrade honestly when things fail.
Memory for LLM apps: extraction, deduplication, decay grounded, not vibes.
Evals as a CI gate, model comparison matrices, latency/cost trade-offs.
Guardrails & red-team basics: fail-closed auth, prompt-injection defenses.
Make AI-generated code trustworthy: compile-verify, property-test, or roll it back.
🚪 Backend engineer moving into AI?
I made that pivot myself I'll show you the map, not the brochure.
What to learn, in what order, and what to skip.
Build a portfolio that proves it (mine: 4 open-source AI projects, one on Maven Central).
🎯 My Teaching Style
No ego, no jargon → Ask anything. I'll meet you where you are.
"Why" over "what" → I'll help you understand, not just repeat.
We build on YOUR real project → not toy examples.
Adaptable & fun → if cat memes help you learn, I'll bring the cats. 🐱
✅ Bottom line?
Whether you're demoing, shipping, or pivoting we'll make your AI actually work, one clear concept at a time.

