Phase 10: AI Engineering Career System · 70 min · GitHub · Docker · Loom
Capstone & Demo Day
Your capstone is the one project that proves you can build, evaluate, ship, and explain.
Hiring signal: End-to-end ownership of a production-style AI system
What you will learn
- Scope a capstone that integrates multiple course skills in 2 weeks
- Assemble the universal deliverables hiring managers look for
- Produce an eval report and a cost report, not just a demo
- Record a tight demo video that shows value and reliability
- Defend the system's tradeoffs in a demo-day Q&A
The Problem
A portfolio of small projects shows range. A capstone shows depth — that you can take one real problem all the way from idea to a deployed, evaluated, explained system, the way you'd have to on the job. It's the single artifact most likely to anchor your interviews: "tell me about your capstone" is a question you want.
Most capstones fail in one of two ways: they're a flashy demo with no evaluation (so reliability is unknown), or they're technically deep but undemoable (so no one understands the value). This lesson is about doing both.
The Concept
A strong capstone integrates skills from across the program into one coherent system:
You have 2 weeks for your capstone. You plan to build a multi-agent RAG system with 5 tools, custom fine-tuned reranker, and a React frontend. Is this well-scoped?
The #1 capstone mistake is over-scoping. A working, deployed, evaluated thin slice beats an ambitious half-finished project every time. Week 1: ugly end-to-end baseline (RAG with off-the-shelf components) + first eval number. Week 2: ONE meaningful improvement (measured), deploy, polish, write-up. Interviewers want to see judgment, not ambition.
Pick ONE track (e.g., RAG study assistant, research agent, knowledge bot) and go deep:
Build : the working system (RAG / agent / model service)
Evaluate : an eval suite with real metrics (faithfulness, task success, etc.)
Deploy : a live, reachable demo (container + cloud host)
Explain : architecture diagram, case study, demo video
Account : cost report and security/safety notes
The differentiator between a student project and a hireable one is the last three: evaluation, deployment, and explanation.
The universal deliverables
Every capstone, regardless of track, ships these:
1. Public GitHub repo with a strong README
2. Live demo (a URL someone can click)
3. Architecture diagram
4. Eval report (metrics + before/after of one improvement)
5. Cost report (cost per request/session, and how you'd reduce it)
6. Security / safety notes (threats considered, guardrails)
7. Demo video (2-3 min)
8. Case study (the story: problem, approach, tradeoffs, results)
9. Resume bullets generated from the above
Unlock the full lesson
You've read the first 2 sections. The rest of this lesson covers Build It, Evaluation, Cost, Capstone Scope (fill this out BEFORE coding), Live Demo, What It Does, Architecture, Evaluation, Cost, Tech Stack, Running Locally, Key Tradeoffs, Use It, Ship It, Exercises, Key Terms, Common Pitfalls, Interview Framing — plus a hands-on lab, quiz, and project artifact.
Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.
Browse all courses · View pricing · DeVenture Academy