Data Science Practice Coach

2. Content Generation

Based on the selection, read the corresponding reference file and generate content:

Difficulty mapping:

• 🟢 Beginner: single concept, guided hints, simple data
• 🟡 Intermediate: combine 2-3 concepts, real-world-like data, minimal hints
• 🔴 Advanced: multi-step, edge cases, performance considerations, no hints

Adaptive difficulty: Check progress. If the user's avg_score > 0.85 on the selected topic at the selected difficulty, suggest leveling up. If < 0.6, suggest stepping back.

3. Exercise Delivery

For coding exercises:

1. Present the problem statement clearly.
2. Provide a starter code block with # YOUR CODE HERE placeholders.
3. Include sample input/output.
4. After the user writes code, run python <skill-dir>/scripts/grader.py with the user's code and expected output to auto-grade.
5. Then review the code yourself for style, efficiency, and best practices.

For concept quizzes:

1. Present questions one at a time or all at once (user preference).
2. After answers, score and explain each correct answer.

For mini-projects:

1. Run python <skill-dir>/scripts/scaffold_project.py --topic <topic> --difficulty <diff> to generate the project skeleton.
2. Walk the user through it step by step.

4. Grading & Feedback

After each session:

1. Score the work (0-5 scale for exercises, percentage for quizzes).
2. Identify weak areas from mistakes.
3. Run python <skill-dir>/scripts/progress.py --update --topic <topic> --format <format> --difficulty <diff> --score <score> --max-score <max> --weak "<area1>,<area2>" to persist progress.
4. Show updated stats.

5. Spaced Repetition

Run python <skill-dir>/scripts/progress.py --review-queue to get items due for review. Prioritize these in the next session's exercise selection.

Question Bank

The skill includes a question bank at <skill-dir>/references/question-bank.json containing curated exercises. When generating exercises:

1. First, try to draw from the question bank using:

   python <skill-dir>/scripts/question_bank.py --topic <topic> --difficulty <diff> --n <count>
   

2. The selector automatically prioritizes questions from the user's review queue (spaced repetition).
3. If the bank has no matching questions (e.g., all completed), generate fresh exercises dynamically using the reference guides.
4. After grading, record question_ids in the session for deduplication.

Importing from Existing Notebooks

The user has existing 500-question notebooks for pandas and scikit-learn. To import:

python <skill-dir>/scripts/import_notebook.py --input pandas_500Q.ipynb --topic pandas --append-to <skill-dir>/references/question-bank.json
python <skill-dir>/scripts/import_notebook.py --input sklearn_500Q.ipynb --topic sklearn --append-to <skill-dir>/references/question-bank.json

Important Notes

• Always read the relevant reference file BEFORE generating exercises to ensure accuracy.
• Coding exercises must have deterministic, testable outputs.
• For deep learning: focus on architecture understanding and code reading, not actual training.
• Keep exercises practical — use real-world-like scenarios (clinical data, time series, NLP).
• Progress file location: ~/.datascience-skill/progress.json