Micro‑Foundations (10 days, 30–45 min/day)

Exit checks: vectorize a transform; explain cosine similarity; stratified splits; read a confusion matrix; implement top‑k; CI green.

• F1 — NumPy muscle: vectorization; implement minmax_scale(X) + test.
• F2 — Linear algebra: dot/matmul; cosine_sim_matrix(A,B).
• F3 — Calculus: gradient of MSE; 15‑line gradient descent demo.
• F4 — Stats: bootstrap 95% CI for accuracy (1k resamples).
• F5 — DSA: heapq.nlargest top‑k vs sort; Big‑O sanity.
• F6 — Splits: stratified train/val/test helper with fixed seed.
• F7 — Metrics: accuracy/F1; confusion matrix plot helper.
• F8 — Regularization: LogReg L2; show val improvement.
• F9 — Testing & CI: 3–4 pytest cases; keep main green.
• F10 — Synthesis: learning curve cell + 3 insights bullets.