Independent · applied AI
AI Job-Application Copilot
Tailors résumé bullets to a job — and grades itself.
- Year
- 2026
- Role
- Solo · in progress
- Status
- In progress
- Type
- Independent
An LLM + embeddings tool that rewrites résumé bullets against a specific job description, with an evaluation harness that scores its own output.
The problem
Most “AI résumé” tools are a single prompt with no idea whether the output is good. The interesting engineering is the part everyone skips: retrieval over your real experience, and an eval harness that measures relevance instead of vibes.
Approach
- Embed a structured experience bank; retrieve the most relevant evidence per job description before generation.
- Generate tailored bullets with the Claude API, grounded only in retrieved facts to avoid fabrication.
- Ship a golden-set eval harness (relevance + faithfulness scoring) so changes are measured, not guessed.
What it proves
- Will demonstrate real AI engineering: retrieval, grounding, and evaluation — dogfooded on my own job hunt.
Stack
App
Next.jsTypeScript
AI
Claude APIEmbeddingspgvector
Data
PostgreSQLEval harness