Live demo · runs in the browser
AI résumé builder with an ATS score
Scores a résumé from 0 to 100 against a job description, lists the keywords it is missing and drafts a summary with Claude. React, TypeScript and Zustand, with no backend.

How the ATS score adds up to 100
A heuristic, as the app itself says: it does not promise that every ATS will parse a résumé the same way.
- Section coverage
Experience, education, skills and projects are all there
20 - Action verbs
Bullet points start with what you did
20 - Quantified results
Numbers in the bullets, not only duties
20 - Contact and keywords
Contact details, and the job’s keywords found in the text
15 + 15 - Format and length
A length and layout an ATS can read
10

How I use models
I treat a model’s output as a draft: Sleep Loop waits for a person to approve before it merges, and Gridwise checks an LLM’s plan against hard limits before a solver runs it.
AI harness and LLM projects on GitHub
The harness is the part around the model: its loop, tools, memory, checks and the gate a person signs off.
- Sleep Loop
Repairs code overnight with maker and checker agents, memory across failures and a human approval gate
Python - Gridwise
Turns an operator’s notes into checked rules, then plans the cheapest day of campus energy
Python - Claude Code workflow
A slash command sends a subagent to research, then a skill formats what it found
Markdown
