One AI engineering concept per day. One working build per day, committed to your public GitHub. Two live sessions per week with Himanshu.
Not a course. Not a certificate. A thirty-day sprint that produces a public commit history, an interview question bank you wrote yourself, and a GitHub profile that answers the question "can you actually build this" before anyone has to ask it.
Most AI courses hand you finished code and a video of someone else typing. You watch, you nod, you forget by Thursday. Build30 inverts that: you get the concept in ten minutes, then you write the code yourself in forty-five, on a fresh file, with your own bugs. The struggle is the point. Nobody hires you for having watched a video.
10 min Learning Manual, read the concept
45 min The Daily Build, write the code and commit it
15 min Problem Statement, attempt it
8 min Interview Questions, read both and write your answers
7 min AI-DSA Primitive + Software Design Pattern, read them
5 min Commit to GitHub
─────────────────────────────────
90 min Done
Two days a week you add an hour for the live session. Saturdays are Ship Day, twenty minutes. Sundays are Read Day, thirty minutes.
Nothing takes four hours. If a day takes you three hours, the day is designed wrong, not you.
Week 1: How AI Actually Works Tokens, embeddings, vector stores, and your first RAG system built from an empty file.
Week 2: Building With AI Chunking, hybrid search, structured outputs, tool calling, and your first working agent.
Week 3: Making It Production Ready Agents, evals, guardrails, FastAPI, Docker. The gap between "works on my laptop" and "someone else can use it."
Week 4: The Forward Deployed Engineer Layer System design, client communication, portfolio packaging, and how to find your first AI engineering role or client.
Harvard's mlsysbook.ai covers similar ground under "ML Systems Engineering," self-paced and textbook-style. Build30 differs on three points: it maps every concept to a DSA/interview bridge, it scopes strictly to inference and serving (no training), and it runs as a live cohort with a 30-day deadline instead of an open-ended read.
Day Zero - Start Here for who this is for, or straight to 30 Day Roadmap for the full day-by-day schedule.