The on-ramp · 12 weeks · free
From third semester to AI builder. One focused term.
A free 12 week path. Curated by the best people on the internet. One shipped GitHub repo per week. No videos to passively watch, no buffet of options, no roadmap to hop. Hinglish bilkul welcome hai.
Honest framing
DSA prep is for the interview. Foundations is for the job.
Striver and Apna get you past the screening round. Foundations is what you wish you knew the moment you joined a real AI team and someone asked you to ship something by Friday. It bolts on; it does not compete.
The path
Twelve weeks. One repo each. Two checkpoints. One final gate.
Read the one curated resource for the week, build the artifact, ship the repo. Week 4 is a coach-off micro-checkpoint where you reproduce the prior week from a blank file. Week 12 is the final ship gate: AI off, 3 hour window, deploy something real.
- Week 1: Ship something on the internet today Start here Stand up a public identity (GitHub account, git basics), open Google Cloud Shell in the browser, and deploy a one-page personal site to a live Cloud Run URL inside session 1.
- Week 2: Linux, the command line, and Python that actually runs Be comfortable in a terminal, manage Python with uv or venv, install packages, run a script on Colab and locally.
- Week 3: Python you will actually use to build AI systems Move from toy Python to idiomatic Python (dataclasses, comprehensions, JSON, venv, reading library docs).
- Week 4: Numbers that move: NumPy and vectorized thinking Coach-off checkpoint Internalize broadcasting, indexing, matrix ops at CS231n day-one bar.
- Week 5: Just enough math to not bounce off Karpathy Visual intuition for derivatives, chain rule, vectors, matrices, the Gaussian.
- Week 6: A neural net from scratch, no framework magic Follow Karpathy micrograd, rebuild autograd from first principles, train a tiny MLP, explain backprop in your own words.
- Week 7: A real model, deployed for real Train an image classifier with fast.ai on Colab and deploy it as a live public Gradio app on us-central1 Cloud Run (NOT HuggingFace Spaces).
- Week 8: HTTP, JSON, and reading other peoples APIs Coach-off checkpoint Internalize how the web actually talks; consume a real public REST API from Python without breaking on errors.
- Week 9: Calling an LLM like an engineer, not a tourist Call a real LLM API from Python, manage prompts and responses as structured data, secrets via env, log every call.
- Week 10: Tools, schemas, and the discipline behind tool use Read Anthropic Building Effective Agents end to end, then implement a single tool-using LLM call where the tool is a typed function selected via JSON schema.
- Week 11: Deploy an agent the world can hit Containerize and deploy the Week 10 tool-using agent as a public HTTPS endpoint on free us-central1 Cloud Run with /health and structured logs.
- Week 12: Readiness Checkpoint Final ship gate Prove transfer: in a single coach-LOCKED 3-hour window, build and deploy a fresh tool-using agent against a real public REST API the student has not used before, push public, write a one-paragraph reflection.
No promises we will not keep
What this is, what it is not.
We measure shipped artifacts, not lessons completed. The CS50 baseline says 89 percent of engaged learners never even submit pset0. Our promise is a real portfolio and a real on-ramp into the Agentic Systems Track, not a placement guarantee. We will publish our actual registered to shipped numbers after the first 200 learners finish Week 1, raw, no curation.
Where this leads.
Foundations is the on-ramp into the Agentic Systems Track. Finish the twelve weeks with a public repo for each one, and you walk into that track with the muscle memory of someone who ships every Sunday. Skip Foundations and you can still take the elite tracks; you will just feel the gap. That is your choice to make, honestly.
See the 8 elite tracks