| 📊 View Lecture Slides | Full-screen presentation with navigation |
Session 15: AI Product Lab & Showcase
| Session Duration: 2 Hours | Block: 3 — Agents, Evaluation & Deployment |
Session clock
| Minutes | Mode | Focus |
|---|---|---|
| 0–45 | Prep | Finalize Demos, Diagram, & Deliverable Gates |
| 45–120 | Showcase | Live Presentations (7 min + 3 min Q&A per team) |
Learning Objectives
By the end of this session, students will be able to:
- Deliver a structured, professional technical product presentation to a mixed audience of peers and instructors.
- Demonstrate a working AI-powered web prototype end-to-end, proving the integration of UI, Express Backend, RAG, and the Gemini API.
- Articulate system architecture, engineering design decisions, and quantifiable evaluation results.
- Reflect critically on the limitations of current AI models and outline realistic next steps for product development.
Hour 1: Preparing Your Showcase (45 minutes)
The first 45 minutes of this session are dedicated to final polish. No new features should be coded during this time. Focus entirely on ensuring your prototype runs stably on localhost and your presentation narrative is clear.
1.1 The Presentation Narrative
Your 7-minute presentation must follow this specific arc:
- The Problem: What specific, painful problem does your AI product solve? Who is the user?
- Live Demo (The “Wow” Moment): Show the product working immediately. Run three “known-good” (Happy Path) inputs to prove functionality.
- Honest Failure: Demonstrate one input that breaks the system or causes a hallucination. Explain why it fails.
- Architecture: Show your diagram. Explain how the data flows from the browser, to Express, to the RAG vault, to Gemini, and back.
- Evaluation & Security: Briefly summarize the findings from your
test-suite.jsonevaluations and yourAI-Risk-Report.md. - Next Steps: If you had 30 more hours, what would you build next?
Pro-Tip: APIs go down. Wi-Fi fails. Always have screenshots or a pre-recorded screen capture of your app working perfectly, ready to show if the live demo crashes.
1.2 The Quality Gates
You cannot present unless these two files exist and are populated in your project repository:
test-suite.json(Minimum 15 scored test cases, from Session 13).AI-Risk-Report.md(Completed risk assessment, from Session 14).
Hour 2: Live Showcase (75 minutes)
The floor is yours. Each team will have exactly 7 minutes to present, followed by 3 minutes of Q&A from the class and instructor.
Evaluation Criteria for the Showcase:
- Functionality: Does the prototype actually work end-to-end without crashing?
- Architecture Rigor: Did the team correctly separate the frontend UI from the Express backend logic? Are API keys secure?
- Evaluation Rigor: Did the team honestly test their system, or did they only present easy questions?
- Responsible AI: Did the team identify credible security and bias risks in their Risk Report?
Course Summary: What You Have Built
Over the last 15 sessions, you have transitioned from a consumer of AI chatbots to an engineer of AI systems.
| Session | What You Built / Learned |
|---|---|
| 1 | Set up an AI-native workspace (VS Code, Copilot, Node, API keys) and defined a product idea. |
| 2 | Mastered structured prompting and created a prompt library. |
| 3 | Built a personal knowledge vault (Exocortex) using Markdown. |
| 4 | Applied Context Engineering templates to format unstructured web data. |
| 5 | Learned AI-assisted software engineering workflows (Explain, Generate, Debug) with Copilot. |
| 6 | Scaffolded a custom HTML UI connected to a mock /query backend. |
| 7 | Replaced the mock with a real integration of the Gemini API in Node.js. |
| 8 | Generated mathematical Vector Embeddings from your Markdown vault. |
| 9 | Wired up a complete Retrieval-Augmented Generation (RAG) pipeline. |
| 10 | Built a chained, multi-step LLM Agent workflow. |
| 11 | Implemented Function/Tool calling, allowing the AI to request actions (Calculator & Search). |
| 12 | Extracted structured JSON data from images using Multimodal pipelines. |
| 13 | Built an automated test suite to measure Correctness and Relevance, and debugged RAG failures. |
| 14 | Implemented security guardrails against Prompt Injection and drafted a professional AI Risk Report. |
| 15 | Demonstrated a complete, working prototype in a technical showcase. |
What Comes Next?
Your prototype is functional, but it is not yet production-ready. To take this product to the next level, consider these next steps:
- Cloud Deployment: Deploy your Express app to a service like Render, Railway, or Google Cloud Run.
- Vector Database: Replace
embeddings.jsonwith a scalable vector database like Pinecone, ChromaDB, or pgvector. - Authentication: Add user login (e.g., using Supabase or Firebase Auth) to restrict who can use your AI and consume your API credits.
- Portfolio: Push your code to GitHub (ensure
.envis ignored!) and write a short technical blog post about your architecture.
Final Reflection Questions
After presentations are complete (or as homework), write short, thoughtful answers to these questions in your vault. This solidifies your learning.
- What was the absolute hardest technical problem you had to solve in this module?
- What is the single most important concept you learned about building AI applications that you did not know 15 weeks ago?
- If you were granted another 30 hours to work on this exact product, what specific feature would you build next?
- Where did the AI tools (Copilot, Gemini) fail you the most, and how did you compensate for that failure as an engineer?
Final Deliverables Checklist
Before concluding the course, ensure the following artifacts are present in your project:
- A working Express application with the Gemini API called from the server (not the browser).
- A RAG pipeline successfully retrieving context from at least 10 Markdown documents.
- At least one functioning Tool Call (e.g., Calculator or Search).
- A
test-suite.jsoncontaining ≥15 cases with documented manual scores. - A completed
AI-Risk-Report.md. - An Architecture diagram (Excalidraw or image file).
- Your prepared presentation slide deck or notes.
Congratulations on completing the AI Product Engineering course!
