AI projects, specified properly
Not a list of ideas. Every project carries the architecture, the build in order, how to tell whether it works, what breaks on real input, the deployment path and the repository layout — plus how to explain it in an interview.
12 projects · 3 readable in full without paying · new ones added to the Vault
PDF Question-Answering Agent
Ask questions of a folder of PDFs and get answers with the page they came from.
Tool-Using Research Agent
An agent that searches, reads, and produces a cited summary — stopping when it has enough.
An MCP Server Over Your Own Documentation
Expose your docs as tools any MCP client can use — the read-only first server to build.
Support Ticket Triage and Routing
Classify incoming tickets, extract the details that matter, and route with a confidence threshold.
Read-Only SQL Agent
Answer questions about a database in natural language, without letting anything write.
A RAG Evaluation Harness
Measure retrieval and generation separately, so you know which one to fix.
Meeting Transcript to Tracked Actions
Turn a transcript into owned, dated action items — and nothing else.
Invoice Extraction with a Confidence Gate
Pull structured fields from invoices, and know which ones need a human.
Multi-Agent Content Pipeline with a Critic
Research, draft and critique as separate agents — and prove the critic earns its cost.
Semantic Search Over a Codebase
Find code by what it does rather than what it is called.
LLM Cost and Latency Observability
Know what every feature costs per user before the invoice tells you.
Resume and Job Description Matcher
Score a CV against a role with reasons, not a black-box percentage.
Want the full specifications?
The Vault opens every project: build steps in order, evaluation, failure modes, deployment and repository layout. One payment, permanent access, and new projects are added to it.