Ground generation in candidate facts
Resume and cover-letter generation uses approved profile information, with prompt-injection defenses.
Applied AI / Live beta
The Job Pursuit
From scattered job listings to a reviewed application pipeline.
The problem
Finding relevant roles involves repeated collection, eligibility checks and document preparation. The Job Pursuit brings those steps into a single workflow.
What I built
Built a Python and React application combining employer-source ingestion, eligibility and ranking, grounded document generation and candidate-reviewed application assistance.
Architecture at a glance
A simplified view of the main application components.
- 01
Sources
Employer ATS listings including Greenhouse, Lever and Ashby
- 02
Preparation
Ingestion → deduplication → eligibility checks
- 03
Matching
Embeddings and LLM-assisted ranking
- 04
Delivery
Grounded documents → candidate review → application tracking
Engineering decisions
Keep the candidate in control
Submission assistance pauses for confirmation and unsupported or ambiguous steps.
Separate collection from decision-making
Ingestion and eligibility checks prepare listings before ranking and document generation.
Technology
Current outcome
Deployed beta with a pipeline that has ingested and scored more than 10,500 postings across five sources.
What I would improve next
The next evaluation should compare ranking quality against candidate judgments and measure document factual accuracy.