Sprint Calendar

S1 through S26 — The Complete 52-Week Schedule

Each sprint is exactly 2 weeks. 4 sprints are reserved as buffer weeks — they appear in the table but carry no new learning objectives. Slippage in any sprint should be logged in the template at the bottom of this file, not ignored.

Starting date: Monday, July 28, 2026. Adjust from your actual start date if different — the structure is what matters, not the calendar dates.

The rule on buffer sprints: they absorb slip. They do not absorb new content. If you find yourself “ahead” on a buffer sprint, use it to consolidate, review lab notebooks, or rest. Never use buffer time to accelerate into the next phase.

---

The Sprint Table

Sprint	Weeks	Dates (approx)	Phase	Primary Focus	Deliverable	Lab Notebook Expectation	Load
S1	W1-2	Jul 28 – Aug 10	Ph0	Linear algebra: vectors, matrices, operations	Linear algebra exercises complete	2 entries: predict + measure on a concept	Medium
S2	W3-4	Aug 11 – Aug 24	Ph0	Calculus, optimization, chain rule	Gradient descent in NumPy on quadratic	2 entries	Medium
S3	W5-6	Aug 25 – Sep 7	Ph0	Probability, Bayes, distributions	Probability derivations notebook	2 entries	Medium
S4	W7-8	Sep 8 – Sep 21	Ph0	Information theory + BUFFER WEEK W8	Rung 1 SHIPPED (SVD + gradient descent GitHub)	1 entry + sprint retro	Low-Medium
S5	W9-10	Sep 22 – Oct 5	Ph1	Supervised learning: regression, classification	sklearn exercises on 2 real datasets	2 entries	Medium
S6	W11-12	Oct 6 – Oct 19	Ph1	Unsupervised learning: clustering, PCA, t-SNE	Clustering + PCA notebook with all 5 t-SNE failure modes demonstrated	2 entries	Medium
S7	W13-14	Oct 20 – Nov 2	Ph1	Model selection, evaluation, feature engineering	Full evaluation pipeline with ROC/PR/calibration	2 entries	High
S8	W15-16	Nov 3 – Nov 16	Ph1	Classical ML in production + BUFFER WEEK W16	Rung 2 SHIPPED (3-model comparison, GitHub)	1 entry + sprint retro	Medium
S9	W17-18	Nov 17 – Nov 30	Ph2	Neural networks from scratch, backpropagation	Backprop in NumPy, XOR verified	2 entries	High
S10	W19-20	Dec 1 – Dec 14	Ph2	PyTorch fundamentals, training loop, autograd	Full PyTorch training loop on MNIST	2 entries	High
S11	W21-22	Dec 15 – Dec 28	Ph2	CNNs + sequence models (RNN, LSTM, GRU)	CNN on CIFAR-10 >85%	2 entries	High
S12	W23-24	Dec 29 – Jan 11	Ph2	Training dynamics + BUFFER WEEK W22	Rung 3 SHIPPED (backprop + CNN + CharLM)	1 entry + sprint retro	Medium
S13	W25-26	Jan 12 – Jan 25	Ph3	Attention mechanism, transformer architecture	Self-attention from scratch, correct shapes	2 entries	High
S14	W27-28	Jan 26 – Feb 8	Ph3	BERT, GPT family, HuggingFace ecosystem	nanoGPT trained on small corpus	2 entries	High
S15	W29-30	Feb 9 – Feb 22	Ph3	Vision Transformers (ViT), CLIP	ViT conceptual implementation, ViT vs CNN comparison	2 entries	High
S16	W31-32	Feb 23 – Mar 8	Ph3	Diffusion models + GNNs	Rung 4 SHIPPED (nanoGPT + DDPM MNIST + GCN Cora)	1 entry + sprint retro	High
S17	W33-34	Mar 9 – Mar 22	Ph4	RLHF, InstructGPT, DPO derivation	DPO loss derived from scratch, written explanation	2 entries	High
S18	W35-36	Mar 23 – Apr 5	Ph4	PEFT, LoRA, QLoRA implementation	QLoRA fine-tune running (even rough)	2 entries	High
S19	W37-38	Apr 6 – Apr 19	Ph4	RAG systems: retrieval, indexing, evaluation	RAG pipeline on a real corpus, RAGAS baseline	2 entries	Very High
S20	W39-40	Apr 20 – May 3	Ph4	LLM agents, tool use, multimodal	Rung 5 SHIPPED (ALL 4 artifacts required — GATE)	1 entry + sprint retro	Very High
S21	W41-42	May 4 – May 17	Ph5	Experiment tracking: MLflow, W&B, DVC	MLflow tracking 20+ experiments	2 entries	Medium
S22	W43-44	May 18 – May 31	Ph5	Model serving: FastAPI, vLLM, PagedAttention	vLLM latency benchmark vs HF baseline	2 entries	High
S23	W45-46	Jun 1 – Jun 14	Ph5	Containerization, Docker, CI/CD	Docker image built, GitHub Actions pipeline green	2 entries	High
S24	W47-48	Jun 15 – Jun 28	Ph5	Monitoring, data quality, Evidently	Rung 6 SHIPPED (live system, GATE) + Rung 7 SHIPPED (paper reproduction)	1 entry + sprint retro	Very High
S25	W49-50	Jun 29 – Jul 12	Ph6	Paper reading system, frontier papers, OSS	First OSS PR merged, 3 papers annotated	2 entries	Medium
S26	W51-52	Jul 13 – Jul 27	Ph6	Technical identity, pitch, BUFFER WEEK W52	Rung 8 SHIPPED (3 posts + talk/review submitted)	Final sprint retro + full year review	Low

---

Buffer Sprints Summary

Buffer	Sprint	Week	Purpose
Buffer 1	S4 (W8)	Sep 14-21	Phase 0 consolidation, Rung 1 polish
Buffer 2	S8 (W16)	Nov 10-16	Phase 1 consolidation, Rung 2 polish
Buffer 3	S12 (W22)	Dec 22-28	Phase 2 consolidation, Rung 3 polish
Buffer 4	S26 (W52)	Jul 20-27	Final consolidation, pitch rehearsal

Rule: Buffer weeks are single-week pauses within their sprint. The sprint continues — only new material stops.

---

Load Legend

Load Level	Hours/week	Meaning
Low	8-10	Recovery, consolidation
Medium	10-12	Standard learning pace
High	12-14	Active implementation required
Very High	14-15	Hard gate sprint — full effort

---

Slippage Log Template

Copy this block every time you slip. Date it. Be honest. Slippage not logged is slippage repeated.

## Slippage Entry — [Date]

Sprint: S[XX]
Deliverable missed: [What didn't get done]
Reason (one sentence, no excuses): [reason]
Impact on timeline: [0 days / X days]
Recovery plan: [specific action in next sprint]
Will this affect a hard gate? [Yes / No]
If yes — gate contingency: [specific plan]

---

Year-at-a-Glance (Phases by Quarter)

Q3 2026 (Jul–Sep): Phase 0 Math + Phase 1 Classical ML begins
Q4 2026 (Oct–Dec): Phase 1 complete + Phase 2 Deep Learning
Q1 2027 (Jan–Mar): Phase 3 Modern Architectures + Phase 4 GenAI begins
Q2 2027 (Apr–Jun): Phase 4 GenAI complete + Phase 5 Production
Q3 2027 (Jul):     Phase 6 Mastery + Final artifacts + Pitch

---

Return to: README.md · Next: 03_north_star_artifacts.md