AI Career Agent
Autonomous job-scanning & AI-scoring agent
What I built
Sifting job boards for the right roles is slow, repetitive, and easy to do badly. This is a Python agent I wrote that continuously scans job feeds, filters by role and location, removes duplicates, then uses an LLM to score each role against a candidate profile — logging only strong matches to an Airtable pipeline with apply-status tracking. It shows I can architect and ship a working agent in real code, and that I think about cost and reliability while doing it.
The agent pipeline
The engineering behind it
Cost-aware by design
Cheap deterministic rules (keyword + location) run first; the LLM only ever touches candidates that already passed — so spend scales with quality, not volume.
Never runs twice, and resilient
A saved list of hashed "seen keys" means the same job is never processed or logged twice, and retry-with-backoff on the Airtable API keeps a run from failing on a transient error.
Structured, actionable output
The LLM returns a fit score, keep/reject decision, and a reason as clean JSON — every kept role lands in Airtable ready to act on.
Where it's headed
Designed as the next build (not yet shipped): for every strong match, automatically tailor the résumé to each role's specific requirements and draft a custom cover letter — turning the agent from a job finder into an end-to-end apply assistant.
Where it stands
The agent is a working demonstration — running in real code, scanning feeds and logging scored, deduplicated matches to Airtable, with the apply-assistant stage in development.