Course wrap and your capstone from here
Five sessions ago this was a single, unreliable call to a model. It is now a system: grounded in your own data, capable of planning and using tools, checked against real failures instead of vibes, and able to remember across a session instead of starting blank every time. That arc - reasoning, RAG, agents, evals, memory - is the whole shape of production AI engineering. Everything else is depth on one of those five layers.
What actually carries forward
Reliability thinking
Turning a model call into observable, guarded software - the Week 1 habit that makes every later layer safer to build.
Grounding
Knowing when an answer needs to come from your own data, and how retrieval quality bounds everything downstream of it.
Judgment about autonomy
Workflow vs agent is a decision, not a default. You now have the vocabulary and the raw-loop understanding to make it deliberately.
TRACE
Trace, Read, Analyze, Codify, Enforce - the loop that replaces "it seems fine" with evidence, on any system, with any vendor's tooling.
Memory and skills
The difference between a system that starts fresh every session and one that compounds, and the write/retrieve/forget discipline that keeps that safe.
Where your capstone goes from here
The build sprint is not new teaching. It is two weeks to close gaps you already know about: harden the eval suite, tighten the security mitigations from Week 3, add the observability you would want if this were actually somebody's job on the line, and make sure the memory store survives a stranger clicking around it. Treat Demo Day as a floor, not a ceiling. The version people see is rarely the last version you will build.
Beyond Demo Day
- Keep the capstone running. A dead URL six months later is a worse portfolio signal than a small, honestly-scoped project that still works.
- Reuse the TRACE habit on the next system you touch, at work or otherwise. It is not specific to this course's stack.
- Stay in the cohort community. The people who kept building after Demo Day are, unsurprisingly, the ones who kept learning fastest.
- Your certificate is issued on capstone completion and Demo Day showcase; see the Demo Day module for the exact submission bar.
Ready when
One last thing
You did not just finish five sessions of material. You shipped a system across five layers that most engineers never get told how to think about explicitly. That is the actual credential. The Demo Day URL is just the proof.
Go deeper (optional)