TECHNEST Track Curriculum
Track: Artificial Intelligence
Track Goal
By the end of this 12-week track, students — who began the course having only used ChatGPT on the web — will be able to:
- Drive AI assistants (Cursor, Claude Code, Gemini CLI) to build and ship full-stack AI applications without hand-writing code, using natural-language prompts as the primary programming interface.
- Stand up production infrastructure entirely through AI-invoked CLIs and MCP servers — Vercel for hosting, Neon for Postgres + Auth, Vercel Blob for file storage, GitHub for source control, Slack for notifications, Typst for PDF generation.
- Own a live, polished, AI-powered personal website that doubles as their portfolio, AI-clone chatbot, guestbook, blog, contact funnel, and the source-of-truth for their auto-generated CV and cover letter.
Capstone Project Goal
In Weeks 9–12 students form small teams (2–4) and evolve their personal-website codebase into a deployed multi-user AI SaaS MVP on one of three tracks:
- Campus Life — event finder with AI recommendations, roommate matcher, language-exchange app, dorm chore-and-bill splitter.
- Personal Growth — habit coach with weekly AI reviews, reading companion, journaling assistant, workout planner.
- Creative Tools — AI short-story studio, music-prompt playground, comic-panel generator, resume-video script maker.
Every capstone ships to Vercel, uses an LLM API, stores state in Neon, and is demonstrated in a live 5-minute team presentation.
The Delegation Rules (course-wide doctrine)
This track runs on one principle.
Students describe what they want. AI writes, installs, configures, deploys, and commits. Students verify. Students only touch files or tools with their own hands when no CLI, API, MCP server, or Skill can do the job — which, in practice, means signing into third-party consent screens and copy-pasting secrets that only a human can see.
| Operation | AI drives via | Student role |
|---|---|---|
| Provision DB / Auth | Neon MCP, Neon CLI | Verify |
| Deploy + env vars | Vercel CLI, Vercel MCP | Verify |
| Git commit + push | Cursor / Claude Code | Verify |
| Install dependencies | Cursor runs pnpm add | Verify |
| Write / edit source files | AI edits the files | Verify + critique |
| Schema migrations | AI + Drizzle CLI | Verify |
| PDF generation | Typst skill in Claude Code | Verify |
| Slack webhook calls | AI fetches / posts via CLI | Verify |
| OAuth consent screens | — (human required) | Do it |
| API keys from provider sites | — (human required) | Do it |
| DNS / domain records | — (human required) | Do it |
.env.local edits when no CLI exists | — (human required) | Do it, lecture shows the exact line |
Every weekly Hands-On Lab applies these rules. Students will recognise the pattern by Week 2 and own it by Week 4.
8-Week Teaching Plan
The entire teaching phase evolves one progressively-built personal website. Each week grafts a meaningful full-stack feature onto the same codebase, so by Week 8 the student owns a single coherent production-grade product.
Week 1 — AI Developer Toolkit Setup
JOBTopic. Install Cursor, Claude Code, Gemini CLI. Install Skills and MCP servers (Vercel MCP, Neon MCP, Typst skill). Set up GitHub and connect to Cursor. Try Wispr Flow for voice-to-prompt input. Practise the delegation pattern on a throwaway sandbox repo.
Skills. Installing dev tools from their official sources · signing into GitHub from an IDE · installing an MCP server and a Skill · dictating prompts via voice · watching AI call a CLI on your behalf.
Week 2 — Deploy Your Personal Website
JOBFUNTopic. Fork a Vercel personal-portfolio template (Magic Portfolio). Customise content through conversation with Cursor. Ship to production via AI-invoked Vercel CLI. Connect a custom domain if the student has one.
Skills. Forking a template through AI · describing site content in natural language · verifying a Vercel deployment · reading Vercel build logs when something fails.
Week 3 — Add an AI Clone of Yourself
FUNCOOLTopic. Integrate Gemini 2.5 Flash as a floating chat widget. Write a persona system prompt that makes the widget answer in the student's own voice (background, interests, project history).
Skills. Obtaining and storing an API key · system-prompt design as a student's "first serious prompt-engineering task" · streaming responses in a Next.js app · prompt iteration through AI collaboration.
Week 4 — Go Full-Stack: Neon + Auth
JOBLIFETopic. Provision Neon Postgres and Neon Auth via MCP/CLI. Add a guestbook where signed-in visitors can leave messages. Drizzle ORM manages the schema.
Skills. Letting AI provision a real database · understanding (but not hand-writing) Drizzle schema · Google OAuth consent as the archetypal manual step · cross-environment env-var plumbing (local + Vercel).
Week 5 — Image Uploads with Vercel Blob
FUNLIFETopic. Extend the guestbook so visitors can attach images / memes alongside their messages. Use Vercel Blob for storage.
Skills. File-upload UX · secure upload tokens · verifying a binary upload round-trip end-to-end · thinking about blob-storage cost and limits.
Week 6 — Blog System
JOBLIFETopic. Graft a /blog section onto the site using a Vercel blog template as reference. MDX posts, syntax-highlighted code blocks, RSS feed.
Skills. MDX authoring · routing in Next.js App Router · consuming a template's patterns through AI rather than by reading its code · publishing a first post.
Week 7 — Real-Time Slack Notifications
JOBCOOLTopic. Build a contact form. On submission, persist to Neon and fire a Slack webhook that notifies the student in real time.
Skills. Slack App creation and webhook setup · writing a Next.js server action through AI · observability-as-notification · thinking about PII and spam protection.
Week 8 — Typst PDF Automation for Job Applications
JOBLIFETopic. Use the Typst CLI and the Typst skill inside Claude Code to generate a pixel-perfect CV and cover letter. Drive the template data from the same Neon profile table that powers the portfolio — one data source, many outputs.
Skills. Data-driven document generation · driving the Typst skill through a prompt · one-click CV / cover-letter download from the site · shipping a job-search-ready portfolio.
Required Tools / Stack
| Layer | Tool | Added |
|---|---|---|
| IDEs | Cursor (primary), Claude Code, Gemini CLI | W1 |
| Skills / MCP | Vercel MCP, Neon MCP, Typst skill | W1 |
| Voice input (optional) | Wispr Flow | W1 |
| Version control | GitHub, driven through Cursor | W1–2 |
| Web framework | Next.js 14 + Tailwind CSS | W2 |
| Hosting / deploy | Vercel (CLI and MCP) | W2 |
| LLM in-product | Gemini 2.5 Flash | W3 |
| Database + Auth | Neon Postgres + Neon Auth + Drizzle ORM | W4 |
| File storage | Vercel Blob | W5 |
| Content | MDX blog | W6 |
| Notifications | Slack Incoming Webhook / Slack App | W7 |
| Typesetting | Typst CLI + Typst skill | W8 |
Assessment Strategy
- Weekly Assignments — 50%. Each week the student submits (a) an updated live site URL and (b) a screenshot of the Cursor / Claude Code conversation that drove the change. The screenshot requirement reinforces the prompt-first doctrine: the prompt is the student's work.
- Participation — 10%. Attendance, voice in Q&A, helping unblock classmates.
- Capstone — 40%. Functionality, deployment health, team collaboration, live demo.
Learning Milestones
By Week 4, students will be able to:
- Open a Next.js project in Cursor, describe a database-backed feature in plain English, and watch Cursor provision Neon, wire Drizzle, deploy to Vercel, and commit to GitHub — end-to-end — from a single conversation.
- Recognise when a step requires human hands (OAuth consent, DNS, provider-side credentials) and know exactly what to do at those moments.
By Week 8, students will be able to:
- Ship a feature to production in a two-hour class and leave with a live URL to share with friends, family, and recruiters.
- Point a recruiter at a single personal site that proves full-stack AI competence: auth, database, file storage, blog, real-time alerts, and AI-generated CV / cover-letter downloads, all served from one Next.js app.
Capstone Evaluation Criteria
- Functionality — every promised feature works end-to-end in production.
- Code quality — reasonable component structure, no hard-coded secrets; AI wrote the code but the team understood and critiqued it.
- UI / UX — clean, accessible, mobile-friendly.
- Deployment — live on the internet, database migrations applied, error-free logs during the demo.
- Presentation — problem statement, solution, tech stack, live demo, challenges faced.
See the full Capstone Specification for tracks, requirements, timeline, and Demo Day rubric.