Four builds. Each one is a line
on your résumé with a link.
Every course runs live in a small cohort and defaults to no-code tools, then invites you to "rebuild it properly" with Python and APIs. Same outcome, your choice of depth. Self-paced versions open after each course's first cohorts.
Data Storytelling with AI
6 sessionsTake a dataset you actually care about — your Spotify history, city 311 data, sports stats — and produce what analysts actually produce: an interactive dashboard and a published data story. AI does the grunt work; you own the editorial judgment, and you're graded on catching the AI's mistakes.
→ Ships: live dashboard + published data articleAI Research Assistant & Accelerator
8 sessionsBring your real research question — thesis, lab project, or competition entry. Each session applies one AI capability to it: grounded literature scanning, contradiction-finding across sources, hypothesis red-teaming, and writing with integrity. You must catch your AI inventing a citation — it's a graded exercise. Builder track: a deployed RAG engine over your own sources, with a hallucination-rate write-up that reads beautifully in a technical interview.
→ Ships: cited research output + RAG system + disclosure statementThe AI Startup LabFlagship
8 weeksServe a real client — a campus org, local business, nonprofit — or found your own team. Interview users, scope the problem, ship a working AI product, run user testing, harden it, and hand it off with documentation. It ends at a public Demo Day in front of founders, faculty, and recruiters. The full arc below ↓
→ Ships: delivered product + GitHub repo + Demo Day pitchAI Agents & Automation Lab
6 sessionsAudit your own recurring drudgery — email triage, meeting notes, weekly reports — and build an agent that does it. The house rule: your automation must run in production for two weeks, and you present measured time-saved data. "Saved 3 hrs/week with an LLM agent" is a résumé bullet with receipts.
→ Ships: production agent + 2-week run log + impact dataEvery course ends in something deployed, published, or running — not a certificate.
Production logs, hallucination-rate write-ups, time-saved data. Claims you can defend in interviews.
Ship fast with no-code, then rebuild it properly with Python. The project motivates the code.
Your thesis, your job hunt, your annoying weekly tasks — the coursework is your actual life.
Eight weeks from "I have an idea"
to Demo Day.
Inside the flagship: the same arc every real product goes through, compressed and coached.
Discovery & scoping
Interview a real client or real users. Write the problem spec. Decide — honestly — whether AI is even the right solution. Most students have never scoped for a real stakeholder; it's the highest-value lesson in the program.
MVP, fast
Ship the no-code MVP: a Custom GPT or Claude Project grounded in your client's documents, wired into their workflow where needed.
Test with real users, then harden
Watch your actual users break it. Iterate. Map the failure modes, add guardrails, and fix reliability — the unglamorous work that separates demos from products.
Rebuild properly → Demo Day
Coding track rebuilds on Streamlit/Gradio with retrieval and logging. Everyone documents and hands off. Then: public Demo Day — pitch, live demo, Q&A — in front of founders, faculty, and recruiters.
Coming soon: the Career Copilot weekend workshop.
Build your AI Career Copilot — working by Sunday
Not "use ChatGPT on your résumé." You build a reusable pipeline: job posting → requirement analysis → résumé gap check → tailored application → interview plan — plus company research and interview-practice personas. Then you run one real application through it before the weekend ends.
Dates and founding pricing are announced to the early-access list first — tell us you're interested below.
Leave with a working career copilot you'll use on a real application that week.
See exactly how our courses run — same instructors, same build-first format — before committing to one.
No code, no prerequisites — any major, any year. We also run this with career centers and student orgs.
Speak fluent 2026 in your next interview.
RAG, agents, evals — from experience, not articles
You'll explain grounding, retrieval, and guardrails because you built and broke them, not because you read a thread about them.
A portfolio that compounds
Each course adds a live artifact + a "How I Built This" write-up with an AI-use disclosure. By the third one, you have a body of work.
Verification as a professional habit
Every course grades you on catching AI failures. Employers are desperate for people who use AI and know when not to trust it.
Works for every major
Biology, English, engineering, economics — the courses run on your domain's real material. AI fluency plus your major is the actual moat.