Using AI to Learn AI
Interactive maps, course radars and market reports I build with AI while learning AI engineering. Each one is a working study tool, published in case it helps you too.
9 pages, newest first
Hands-on tutorial ·
Hands-on Building a Multimodal Document QA System over Real Annual Reports
Build 10-K Lens: search six real bank 10-Ks (2,942 pages) with text, vision and hybrid retrieval, answer with cited pages and highlighted evidence, and ship it as a web app. Measured on 309 human-labeled questions with confidence intervals. Runs on a laptop for $0.
Start the tutorial
PrivateHands-on · P2 ·
Alexandria: An Autonomous Personalized Book Curator That Learns My Taste - P2 Hands-on
Build, deploy, ship and operate Alexandria step by step: a polite crawler, an AI gateway, Postgres with owner-only functions, GitHub Actions, and a daily book page on GitHub Pages. 33 files, 36 tests, $0 on free tiers.
Start building
System design · P1 ·
Alexandria: An Autonomous Personalized Book Curator That Learns My Taste - P1 System Design
Design for Alexandria, a system that crawls the bookstores I track, shortlists books against my recorded taste, checks reviews, publishes one pick a day on this blog, and learns from my reactions. Includes an AI gateway and a cost comparison of AI APIs, free tiers included. Draft for review; nothing is built yet.
Review the design
Hands-on tutorial ·
Hands-on Building Wikipedia Research Agent By Post Training using SFT and GRPO
A copy-paste tutorial: post-train Qwen3.5-4B into an agent that researches Wikipedia, following Samaya's recipe: Eval, Harness, SFT, GRPO, Ablation. 16 tested Python files, every command, real outputs, about $22 on Tinker.
Start the tutorial
Interactive map ·
Applied AI in Business and Solution Design Atlas
A 14-stage dependency graph for turning business needs into AI solutions and building the agents that automate business workflows, with learning routes through AI Engineering from Scratch, Stanford and DeepLearning.AI (technical).
Explore the Atlas
Interactive map ·
The AI Infrastructure Atlas
A 12-stage dependency graph from GPU fundamentals to cluster-scale inference, the learning path for running open models on your own GPUs, course routes through Stanford and DeepLearning.AI, and 14 hands-on labs.
Explore the Atlas
Market report ·
What Vietnam Hires AI People to Do
Job descriptions from three Vietnamese job sites, read and coded: what employers expect AI-era engineers to do, which business problems they pay to solve, and what it pays.
Read the report
Course radar ·
How Stanford Is Transforming CS in the AI Era
What Stanford's CS department added, renamed and retired for 2026–27, what that says about where AI is heading, the fundamentals to learn first, and curated learning paths.
Read the radar
Course catalog ·
Stanford CS Full Courses 2026–27
Every active Stanford Computer Science course for 2026–27, searchable by title, description, domain and status. The companion catalog to the course radar.
Search the catalog
Interactive map ·
The AI Engineering Atlas
78 core AI engineering concepts, deduplicated from a 20-phase curriculum and wired into one dependency graph. Click any concept to see what to learn first and what it unlocks.
Explore the Atlas