← HomeFrom automating the job hunt
to ending it.
Rolemate started in April 2026 as a tool to help people apply faster. Every customer interview pointed the other way. This is how the product found its real problem, trust, and what changed the direction each time.
April 2026
DiscoveryThe first idea: automate the job hunt.
Rolemate started as an AI career assistant. It read your GitHub and LinkedIn and turned them into targeted applications, referral maps, and tailored resumes. The bet was that the grind of applying could be automated end to end.
Roughly 10 customer discovery interviews.
Pitched to students and early builders. The interviews kept returning one thing: applying more is the problem, not the fix.
May 2026
First stage, first proof3rd place, ClickHouse Demo Night NYC.
Competed among 10 companies including Google, Docker, and RedHat, in front of 200+ VCs, angel investors, and founders. The concept held on stage, and the discovery signal was getting louder.
June 2026
The build begins, and the pivotJoined an NSF customer-discovery cohort for the summer.
A summer of NSF-backed customer discovery, run with Entrant. Weeks of interviews sharpened the read on what the market actually lacked.
The career assistant was fully built.
Coach chat, an application-tracking pipeline, resume tailoring, LeetCode practice, bring-your-own API keys, a browser extension. Every piece worked, and every piece already had a good version somewhere else.
The pivot: stop automating the hunt, start proving skill.
The realization: nothing in the career assistant solved a problem advanced builders did not already have a tool for. The one thing missing everywhere was trust, a way to prove what you can actually build. Rolemate became the Signal Score, a verified read of your real GitHub work from a five-agent panel. The look changed with it, from a dark dashboard to warm editorial, with the magenta and indigo from the logo.
Two sides, one score.
Candidates free, businesses paying for verified talent. GitHub OAuth connect, live scoring on real code, and honest Live and Preview labels so the product never overstates itself.
Invited to present Breathe at StartupGrind @ a16z.
Selected among 8 presenters for New York Tech Week, in front of 100+ VCs, angels, and founders. Breathe, the event-intelligence layer shown there, is now Rolemate’s events feature.
A correction: your score is yours.
Shareable public scorecards shipped, then within a day every score became private and account-owned. Candidate proof is a growth engine, and confidential-by-default is the only version of it worth trusting. Data discipline became a founding rule, not an afterthought.
Events, mapped by signal.
The Signal Map: a bubble map that ranks the rooms worth walking into by who is actually there. The score stopped being an endpoint and started routing you toward people.
July 2026
The real visionBigger than GitHub: proof of skill itself.
Reframed from a GitHub scorer to proof-of-skill HR technology. GitHub is the first evidence source. How you work with AI agents comes next: how you prompt, how you catch and fix hallucinations, how efficiently you get to a result. The category is verified proof of skill from real work.
The canon: stop applying entirely.
Zero-application hiring via MCP. Plug Rolemate into the AI agent you already use, and it scores your fit for open roles from your real work, private data never leaving to companies. Cross a company’s competence threshold for a role and both sides are notified, pipelining you straight to an interview. Shipped the same day: the MCP server, People v1 for who to meet, and a split-screen film that shows the applicant building while the same person climbs the employer’s shortlist.
Endorsed by 434 VC.
External conviction in the wedge: verified proof of skill as the trust layer employers reach for before the broken funnel.
The through line: the bottleneck in hiring is trust, and proof of real work is the only thing that fixes it. Everything since April has been moving toward proof you own and companies believe.