Job searching is repetitive and manual: scanning postings, judging fit, rewriting resumes for every role, and cold-emailing recruiters with no idea whether you're even a good match. FirstJob AI automates this entire pipeline — and for the Band of Agents Hackathon, we built the part that makes it genuinely multi-agent: three specialized agents that discover each other, hand off context, and complete real work together through Band's chat-room and @mention system. Job Matcher receives a candidate profile and a job posting, scores the fit (0-100) with specific reasoning, and decides whether to proceed. If the match is strong, it @mentions the Resume Tailor directly in the Band room, handing off the job requirements and match analysis. Resume Tailor picks this up automatically, produces concrete resume edits — reordered skills, a targeted summary, quantified project bullets — then @mentions the Outreach agent with everything it needs. Outreach finds a plausible recruiter contact and drafts a complete, personalized cold email, ready to send. This isn't a linear script calling three functions. Each agent is an independent Band-connected process with its own role, its own reasoning, and the autonomy to decide when to hand off and when to stay silent. Band is the actual coordination layer — context flows through shared chat-room messages and @mentions, not through a central orchestrator dictating each step. FirstJob AI itself is a working full-stack product beyond this hackathon scope: Gmail OAuth, live job scraping with DB-cached match scoring, automated resume tailoring with safety checks, and Hunter.io-powered recruiter outreach, built on FastAPI, PostgreSQL, and React. What we demonstrated through Band is the next evolution of that system — moving from one model doing everything in a single call to a team of focused agents that collaborate the way a real recruiting team would, each excellent at one job, coordinating through a shared platform instead of hardcoded logic.
Category tags: