01-introLearning objective
Intro to Practical ML
Map the end-to-end ML lifecycle and define what success looks like before writing code.
Why this matters
Most beginner projects fail because they skip planning. Great ML starts with clear scope, measurable outcomes, and known constraints.
Visual
ML Lifecycle
+------------+ +------------+ +------------+
| Problem | -> | Data | -> | Features |
+------------+ +------------+ +------------+
| |
v v
+------------+ <- +------------+ <- +------------+
| Deploy | | Evaluate | | Train |
+------------+ +------------+ +------------+Code block (Python)
# quick_project_brief.py
project = {
"problem": "Detect spoiled fruit from phone photos",
"metric": "accuracy",
"target": 0.85,
"latency_budget_ms": 120,
"device": "Raspberry Pi Zero 2 W"
}
for k, v in project.items():
print(f"{k}: {v}")Try it now
python quick_project_brief.pyecho "Ship only what you can measure."
Checkpoint
- I can describe the 7-stage workflow in order.
- I picked one concrete prediction task.
- I selected one primary metric and target value.
Mini quiz
What should be chosen before model training?
- GPU type
- Primary metric
- UI theme
- Cloud provider
Answer: Primary metric
Without a metric you cannot compare experiments objectively.
Which phrase best describes deployment-ready ML?
- Highest benchmark only
- Works within real constraints
- Newest model only
- Most parameters
Answer: Works within real constraints
Real projects require speed, reliability, and reproducibility, not just high score.
Why define latency budget early?
- For better CSS
- To avoid impossible model choices later
- To reduce dataset size only
- For easier naming
Answer: To avoid impossible model choices later
Hardware limits shape model architecture decisions.
Troubleshooting
- If scope feels huge, narrow to one target label first.
- If metric is unclear, choose a baseline metric by task type (accuracy/f1/mae).
- If deployment seems scary, define a local CLI inference as your first deploy target.
Ship it task
Write a 5-line project brief and commit it as `project-brief.md`.
Complete this lesson to unlock redemption eligibility faster.