Tensor-to-Tenant

Build the math, systems, and judgment behind production AI. 108 weeks, three stackable releases, ten gates. The official site and learner cookiecutter stay at GitHub; this track is the same syllabus in AI Engineer Exhaustive’s reading chrome, wired to problems, ML math, LLM internals, and CS/AI 684.

Level: working engineer or DS who can protect 10–15 h/week
Path: tensor → vector → model → serving → platform → tenant

How to use this mirror

1. Open a week. Prerequisites are ordinary links on this site.
2. Do the deliverable in your learner journal.
3. Hit the gate weeks before moving on. Details: Gates.
4. CS/AI 684 is the 14-week deep-dive that sits beside T2T inference weeks (82–93), not a replacement.

How to work a week: Weekly operating model (Mon–Fri loop, Forge, evidence). Engineering stubs: Primitives. Need fundamentals first? 36-week on-ramp. Interview speed lane: Leetcode Darbar (548 catalog pages).

Official browser: https://sethuiyer.github.io/tensor-to-tenant.

Releases

• Foundations (W1–30) — Math, numerical routines, and experimentation toolkit
• Engineering + Systems (W31–69) — Tested primitives, system designs, MLOps tools, and postmortem
• LLM Platform (W70–108) — LLM/RAG work, inference benchmark, platform layer, and capstone

Phases

• Phase 1 — Orientation, tooling, and diagnostics — weeks 1–6
• Phase 2 — Mathematical foundations I: linear algebra & numerical methods — weeks 7–18
• Phase 3 — Mathematical foundations II: calculus, autodiff, probability, statistics — weeks 19–30
• Phase 4 — Engineering micro-projects and applied ML primitives — weeks 31–45
• Phase 5 — System design and distributed systems fundamentals — weeks 46–57
• Phase 6 — ML lifecycle, experimentation, and MLOps — weeks 58–69
• Phase 7 — LLM training, RAG, agents, and evaluation — weeks 70–81
• Phase 8 — LLM inference and performance engineering — weeks 82–93
• Phase 9 — Production AI platform engineering — weeks 94–102
• Phase 10 — Capstone, portfolio, and interview readiness — weeks 103–108

Also on this site

• Problems — the implementation middle T2T already cites
• ML math — phases 2–3
• LLM internals — phases 7–8
• CS/AI 684 — long-context agents (pairs with W82–93)
• Leetcode Darbar — 548-slot timed lane
• 36-week on-ramp — optional beginner course