From notebook to product, from prototype to pitch — teams form, scope locks, and building begins
By the end of this session, you'll have both conceptual understanding and hands-on experience with the foundations of AI.
Teams of 2-3 (or solo) form around a shared problem. Scope is locked by end of class — no more pivoting after today.
Sketch the full system: agent roles, tools, data sources, guardrails, and human review gates. Get feedback from instructor and peers.
Leave class with a working skeleton: agent loop running, at least one tool connected, basic test framework in place.
Outline the 5-minute demo: what you'll show, what questions you'll answer, and how you'll prove it works with live data.
A carefully crafted progression from concepts to hands-on building.
What's expected, how you'll be graded, and what makes a great final project.
Form teams, choose problems, and write a crisp problem statement.
Design your system on paper, then start building the skeleton in Colab.
You have a scope, a team, and a skeleton. The rest is execution.
Dedicated build time with instructor support. Focus on getting the core agent working.
Plan your 5-minute demo and understand what happens next session.
Everything you need to explore, experiment, and build.
All Final Project details for this semester: deliverables, problem ideas, starter resources, and FAQ. Submit your notebook + 1-page brief on Canvas by Tue May 5 at 11:59 PM.
Open project pageReview any session's materials — agent patterns (S6), RAG (S7), multi-agent (S8), industry agents (S9), guardrails (S10), evaluation (S11).
Browse sessionsThe agent design guide you've been reading all semester. Revisit for architecture decisions and pattern selection.
Read guideThree interactive challenges to build your intuition.
The entire session is a structured workshop. By the end, you have a team, a scope, and a working skeleton:
Find teammates (2-3 or solo). Write a 2-sentence problem statement: What does your agent solve? For whom? Why does it matter?
Draw your system: agents, tools, data sources, guardrails, human gates. Get instructor feedback on feasibility and scope.
Get the core agent loop running in Colab. Connect at least 1 real tool. Load sample data. Start test cases.
Dedicated build time. Get the core working. Start your test suite. Use instructor office hours for help.
Outline the 5-min demo: problem → live demo → test results → business case → Q&A. A working demo beats perfect slides.
Scope is king. A simple agent that works perfectly in your demo beats a complex agent that crashes. Lock your scope today and focus on execution.
Build the tests first. Writing test cases before finishing your agent forces you to think about what 'working' actually means for your use case.
Use the full stack. Your project should show what you learned: tool use, ReAct reasoning, domain knowledge, guardrails, and evaluation. That's the whole semester in one demo.
Simple + polished > complex + broken. Demo day rewards agents that work reliably. Cut scope before you cut quality.
Complete these tasks before our next class to prepare for prompt engineering.
Ship it. Live demos, peer voting, and course celebration. Show the world what you built.