Case study · 04
JobGPT — AI job search & application copilot
Overview
Chat-first job search platform that finds roles across major boards, ranks every listing against your resume, evaluates fit with apply/skip guidance, tailors ATS-ready documents, and tracks your pipeline — without leaving the conversation.
Project preview
Technologies used
- Next.js
- React
- TypeScript
- OpenAI APIs
- Career-Ops
- RAG
- REST APIs
Key features
- Natural-language search across Greenhouse, Lever, Ashby, Workable, YC, RemoteOK, USAJobs, and more
- Resume-aware match scores on every result — skills, gaps, and ranked fit instead of keyword noise
- Per-role evaluation with strengths, gaps, compensation signals, and clear apply/skip recommendations
- One-click resume tailoring saved to Documents; evaluation reports and cover-letter workflows from Chat
- Synced surfaces: Job Search board, Reports, Applications pipeline, and Documents library update from chat
- Conversation history, saved searches, and pipeline status (Saved → Applied → Interview) in one workflow
- Built on Career-Ops — open-source job search engine indexing 27k+ listings across 12+ boards
Challenges solved
- Keeping multi-board ingestion reliable when APIs and HTML layouts differ per source
- Scoring fit against a real CV without overfitting to keyword overlap or missing transferable skills
- Syncing chat-driven actions with structured dashboards so users never repeat work across views
- Designing evaluation outputs recruiters and candidates both trust — actionable, not generic LLM prose
Architecture & engineering highlights
Chat orchestrates search, evaluation, tailoring, and tracking. Career-Ops aggregates listings; resume embeddings and LLM tools score fit, generate documents, and write pipeline state to dashboards (Job Search, Reports, Applications, Documents) in real time.
Engineering highlights
Search → score → act
A single prompt like “remote data analyst jobs, 80%+ match” returns ranked cards with match %, skill overlap, gaps, and actions to evaluate, save, or tailor.
Resume-grounded ranking
Every listing is scored against the uploaded CV — surfacing ~94% fits with explicit matched skills (Python, SQL, R) and missing items (e.g. Go).
Full pipeline in one product
From first search through tailored HTML resumes, evaluation reports, and application tracking — four modules stay in sync with Chat as the control plane.
Screenshots
Future improvements
- Public deployment with auth and persistent user workspaces
- Browser extension to evaluate listings from any job board URL
- Interview prep packs generated from evaluation gaps per role
- Team/recruiter view for shared pipelines and hiring manager handoff