Karte 11 · Kapitel capstone
Project design
🟢 USE — Run first
0 - 15 min
Read a real project blueprint
Before designing your own project, study a complete example. This is a project blueprint for a real AI tool built in a previous cohort.
- Your instructor will share the example blueprint on-screen.
- Read it in full. It follows the Project Canvas template.
- Identify: who are the users? What data comes in? What decision happens? What comes out?
- Find the risk section. What did the team identify as the highest-risk component?
- Find the MVP definition. What did they build first and what did they leave out?
- Come prepared with one question about something you would design differently.
Done-Signal: You can describe the example project in one sentence and name its highest-risk component.
🔵 UNDERSTAND — Look inside
15 - 60 min
The Project Canvas
Every capstone project uses this template. Fill it in during today's session with instructor guidance.
👥 Users
Who uses this
Name the specific people who will interact with this system. Not 'researchers' — 'wet-lab postdocs in a protein folding lab who need to compare 20+ papers per week'.
📊 Data
What goes in and out
List the inputs and outputs. Be specific about format and source.
⚠️ Risks
What could go wrong
List: data privacy risks, model failure modes, dependency risks (APIs that could go down), and use risks (what happens if the output is wrong).
🎯 MVP
Minimum viable version
The smallest version that demonstrates the core value. One data source, one output format, one user type. Everything else is post-MVP.
A project that is not constrained is a project that is never finished. The MVP definition is the most important part of the canvas. It forces you to answer: what is the one thing this system must do to be useful? Start there. Everything else is post-MVP.
Probe-Fragen
- Who is the most specific user you can name? The more specific, the better the design.
- What data does your project need that you do not currently have access to? How would you get it?
- If your entire pipeline went down for a week, what would your users do instead? Is your tool a nice-to-have or a must-have?
🟠 BUILD — Make it yours
60 - 90 min
Fill in your Project Canvas
Complete your Project Canvas by the end of this session. You will present it next week.
Aufgabe: Complete the Project Canvas: users, data, risks, MVP, and a rough architecture sketch.
- Download the Project Canvas template from the course materials.
- Fill in Users: two sentences describing your most specific target user.
- Fill in Data: list every input and output with format and source.
- Fill in Risks: at least one privacy risk, one model failure risk, one dependency risk.
- Write your MVP in one sentence: 'The MVP is a system that does [X] for [user] using [data source] and produces [output].'
- Sketch the architecture: 3-5 boxes connected by arrows. Label each box with the tool or service.
- Share your canvas with a classmate for peer review.
Deliverable: Completed Project Canvas shared with your instructor.
✓ SELF-CHECK
Hast du das verstanden?
- I have a specific user, not a generic one.
- I have listed all inputs and outputs with formats.
- I have at least three risks identified.
- I have a one-sentence MVP definition.
The MVP definition is where most projects fail — they try to build too much. What are you most tempted to add that is not in the MVP? Why? What would you lose by leaving it out?
💬 KI-TUTOR
Frag den Tutor zu dieser Karte
Sokratisch: der Tutor antwortet mit Leitfragen statt fertigen Antworten — du erarbeitest die Lösung selbst.
Stell eine erste Frage zu dieser Karte unten.