This week you'll put together a System Readiness Pack — a collection of artifacts that show you can think through what it takes to move from "it works on my laptop" to "it works in the real world." You'll design a system, run a small experiment with uncertainty quantification, and reflect on what you learned.
No GPU needed. Everything runs on your laptop with plain Python and scikit-learn.
| # | Deliverable | File(s) |
|---|---|---|
| 1 | System Design Doc — architecture diagram, components, tradeoffs, and failure modes | docs/system_design.md |
| 2 | Research Implementation — a small experiment with conformal prediction and a fallback decision function | research_impl/run_experiment.py |
| 3 | Experiment Results — what you ran and what you found | research_impl/results.md |
| 4 | Reflection — answer 3 of 5 short prompts | reflection.md |
Plan for roughly 1.5–2 hours:
- ~30 min on the system design doc
- ~45 min on the experiment code
- ~15 min recording results
- ~15 min on reflection
- Install dependencies:
pip install -r research_impl/requirements.txt
- Work through each deliverable at your own pace.
- The starter files have TODOs and guiding prompts — fill them in.
When you're done:
git add -A
git commit -m "Complete Week 24 System Readiness Pack"
git push| Deliverable | What "good" looks like |
|---|---|
| System Design Doc | Architecture diagram is present (Mermaid or image). Components table has 3+ rows. At least 2 tradeoffs with reasoning. At least 3 failure modes with signals and responses. |
| Research Implementation | Script runs end-to-end without errors. Uses a fixed random seed. Produces prediction intervals or sets via conformal prediction. Includes a fallback decision function. |
| Experiment Results | All four sections filled in. Metric is clearly stated. "Surprised me" section shows genuine reflection, not just "it worked." |
| Reflection | 3 of 5 questions answered. Each answer is 2–3 sentences and specific (not generic). Shows connection between the experiment and system design thinking. |