Topnorch

Job hunting is repetitive: read a posting, rewrite a resume, rewrite a cover letter, repeat dozens of times. Blind auto-apply tools spam employers and damage the candidate's reputation.

Built a polyglot backend: a Rust parser for deep PDF resume extraction, Python NLP pipelines that tailor each resume and cover letter to a role, and Go orchestration services managing job dispatch queues, behind a React + TypeScript interface. A human stays in the loop: nothing is sent until the candidate taps apply.

Modular monolith combining compiled Rust utilities, Python inference pipelines and Go orchestration daemons. PostgreSQL with Prisma for candidate profiles, Redis caching, distributed rate limiting to protect candidate accounts across job boards, and Sentry telemetry.

Skill graphs from resumes are matched against parsed job descriptions with vector similarity. Token-bucket rate limiters, shared across workers, keep applications within each board's limits.

Rust where parsing speed matters, Python where the NLP ecosystem lives, Go where concurrency and queues matter: each language sits where its strength is.

Human-in-the-loop automation: the system does the tedious tailoring and matching, and the candidate keeps the final decision.

A career-acceleration layer for job seekers, starting with a waitlist launch.