Case study, 2025
Pathfinder: a multi-agent assistant for the whole job search
Job seekers repeat the same research, tailoring and follow-up for every single application.
What I built
- Specialist agents for resume tailoring, job matching, application drafting and interview coaching.
- An orchestrator coordinates them, so each step hands clean context to the next.
- Personalised retrieval grounds every application in the candidate's real history.
This is confidential client work, so the product name is a stand-in and there are no product screenshots. The architecture and the sample run show how it works.
- Python
- FastAPI
- OpenAI SDK
- Agentic workflows
- Next.js
- PostgreSQL
How it works
Select any node to see what it does.
Illustrative run, simplified
- 00.00orchestratornew target role: senior backend engineer
- 00.18matcher12 openings ranked by fit
- 01.05optimizerresume tailored for the top 3
- 01.47writerapplications drafted, waiting for approval
The architecture, step by step
Each part of the diagram, in the order a run moves through it.
- Candidate historyWhere every run starts.
- OrchestratorRoutes every task to the right agent and keeps the shared context.
- Resume optimizerTailors the resume to each role from the real work history.
- Job matcherRanks open roles by how well they fit.
- Application writerDrafts every application and waits for approval.
- Interview coachPrepares likely questions and gives feedback on answers.
- ToolsThe agents act through the retrieval store, job boards, email and calendar.
Related work
Crewline
A platform that provisions digital employees.
Study Mode
Gemini-powered study sessions built from your own notes.
Knowledge assistant
Answers from company documents, with every claim cited.
Research crew
A team of agents that researches a market and writes the brief.
Evals and tracing
Every agent run traced, scored and compared before it ships.
AI Product Build
A complete AI product: agents, backend, frontend and launch. From $10,000, 8 to 12 weeks.