A working full-stack AI app
Load real documents, chunk and embed them, store them in a vector database and serve answers through an API and a UI. Built by you, with working code at every step.
embeddingsvector DBRAGAPI + UI
ONE-DAY LIVE WORKSHOP // PRESS START
AI Engineer Foundations
One Saturday. Eight hours. You build a full-stack app with RAG, agentic retrieval, evals and observability, and you leave with a plan to build your own.
- Sat, Jan 16, 2027
- 9am to 6pm PT
- Online, live and recorded
- 20 seats
- $497
$10 in OpenAI credits included.
THE SKILLS GAP
Using Claude every day is a fine habit. It won't get you through an AI engineer interview.
I helped design the interview process for an "AI Engineer" role at a startup I worked for. Nobody asked which chatbot you like. They asked how you'd chunk a pile of documents, what you'd do when retrieval comes back with garbage, and how you'd know your agent quietly got worse after a prompt change.
None of this needs a math degree or an ML background. Most experienced developers can learn it. What's hard to find is someone who already made the expensive mistakes and can show you where things break.
That's what this Saturday is for.
WHAT THEY ACTUALLY ASKED ABOUT
- RAG. Chunking, re-ranking and opinions on retrieval strategy.
- Agents and workflows. Tool-calling vs orchestration, and when a framework is worth it.
- Evals and testing. How you prove the agent answers correctly and catch regressions.
- Observability. Token costs, latency and seeing where an agent drifts.
YOUR LOOT
What you walk away with
Agentic retrieval
An agent that writes its own search queries, decides when it has enough context and uses tool-calling with structured outputs, so it does more than stuff text into a prompt.
tool-callingquery planningstructured outputs
Evals and observability
Evals and LLM-as-judge tests that tell you when the agent is wrong, plus tracing so you can see token cost, latency and exactly what the model saw.
evalsLLM-as-judgetracing
A plan for your own project
Before you log off, you map a project from your job or a product idea onto the same architecture. You leave knowing what to build next and in what order.
$10 in OpenAI credits
Enough to build and test everything we cover, so the API bill isn't your excuse.
The full recording
Every minute is recorded. Rewatch the parts that went fast, or catch up if life gets in the way on the day.
THE RUN
How the day goes
Saturday, January 16, 2027, Pacific time. We build one app from an empty folder to something with evals and tracing, in this order.
- Kickoff and setup checkHow LLMs and embeddings work, explained in plain English, and everyone's environment running.
- Data for LLMsLoading documents, chunking strategies, metadata and storing vectors.
- Retrieval that doesn't suckSemantic search, metadata filters, relevance thresholds and re-ranking.
- Lunch45 minutes. Step away from the screen.
- Agentic retrievalTool-calling, letting the agent plan its own queries, structured outputs.
- Make it full-stackWire the agent to an API and a UI people can actually use.
- EvalsTest cases, LLM-as-judge and catching regressions before your users do.
- ObservabilityToken cost, latency and traces. Find out where the agent goes off the rails.
- Your project planMap your own idea onto this architecture and leave with next steps.
- Q&A and wrapDone by 6pm PT.
Planned agenda. Times will flex a little depending on the group's questions.
PLAYER REQUIREMENTS
Is this right for you?
YOU'RE A GOOD FIT IF
- You write code for a living, or you can drive Claude Code or Cursor well enough to ship working code and understand what it wrote.
- You keep hearing RAG, agents and evals and want to build them once, properly, instead of reading another thread about them.
- You built a tutorial chatbot and aren't sure what you actually built.
- You want something real to show for your Saturday.
SKIP THIS ONE IF
- You've never written code. This moves fast and I won't be teaching loops.
- You want a slide deck about the future of AI. We'll be in an editor most of the day.
- You're already shipping RAG pipelines with evals in production. You'll be bored.
YOUR GUIDE
Taught by someone who has shipped this
I'm Brian Jenney. I've built and led production AI systems, including RAG pipelines and agents, at two companies with very different products. Both times I hit the problems you'll hit, usually the expensive way.
I also run Parsity, where we've been training working developers for years. This workshop takes the foundations from our AI engineering program and packs them into one long, hands-on day.
Production RAG and agents at 2 companies
U.S. Department of Labor grant
Official Claude partner
WHY ONLY 20 SEATS
I'll be answering questions and looking at your code all day. Past 20 people, that stops being possible and it turns into a webinar.
When the seats are gone, they're gone.
INSERT COIN
$497
One day, one working app, a plan for your own project, and the recording.
- 8 hours live with Brian, online
- Full-stack app with RAG, agentic retrieval, evals and observability
- A plan to build your own project
- $10 in OpenAI credits
- The full recording
LIMITED TO 20 DEVELOPERS
FAQ
Questions people ask
What if I can't make the whole day live?
Everything is recorded and you get the recording. Live is better because you can ask questions while you build, but you won't lose anything if you have to step away.
Do I need to know AI or machine learning already?
No. You need to know how to code, or be able to use coding agents well enough to build and read real code. We cover the AI parts from the ground up, with no math required.
What do I need on the day?
A laptop with a dev environment you're comfortable in and a decent internet connection. Setup instructions go out before the workshop so we don't burn the morning on installs.
Do I need to pay for API usage?
You get $10 in OpenAI credits, which covers everything we build during the day.
When is lunch?
45 minutes, starting around 12:15pm PT. The day runs 9am to 6pm PT.
Can my company pay for this?
Stripe sends a receipt after checkout that you can submit for a learning budget or reimbursement.