Our mission
Measure what the work actually needs
Technical hiring has two breakdowns we want to solve.
The first is that coding tests have collapsed into a cheating arms race: capable AI agents and screen overlays make take-homes and live-coding screens trivial to pass without exercising the underlying skill. The second is that the capability that does matter, engineering effectiveness when working alongside AI, is not measured anywhere.
Banning AI from the interview selects for unaided coding while leaving the AI-orchestration skill entirely untested. We do something different. We give every candidate a unique conversation built from their own CV, then put them in a real codebase alongside the AI tool of their choice and watch how they ship.
We do not believe AI should make the hiring decision. We believe AI should produce evidence that is quote-grounded, auditable, and fair, so the human interviewer can make a better one.
What we’re building
Two products, one rubric
However you run your process, Basanite plugs in. Hand a round to the agent, or keep your own interviewers and let the co-pilot ride along.
Basanite agent
Runs a hiring round for you.
An adaptive voice interview built from each candidate’s own CV, followed by a written, quote-grounded briefing on their strengths, limits, and technical depth before you ever meet them. Round two puts them in the AI Collaboration Workbench: a sandboxed VS Code environment where they ship a real ticket alongside the AI tool of their choice.
Basanite co-pilot
Sits with your interviewers, live.
In your own rounds, the co-pilot suggests what to probe next in real time, then writes up the session and ranks each candidate with the evidence attached. Your interviewers stay in control; the note-taking and the scoring rigour come for free.
Both are grounded in the same eight metacognitive dimensions and scored on one rubric, every time, so every candidate is read the same way and every number opens to the quote behind it. See the methodology
What we believe
Four principles behind the product
Depth over breadth
Each layer of the assessment exists to move one level deeper into signal quality. A candidate who answers fluently at the surface should encounter ground that shifts beneath them at the next layer. The edges of real ability are blurry; performed ability has no edges.
Structure as fairness
By anchoring every evaluation to consistent frameworks and explicit scoring criteria, a self-taught engineer without institutional pedigree can be seen as clearly as one from a target university. Both are asked the same questions in the same spirit, with the same depth of follow-up.
Honest about AI limits
We flag where human expertise is required, produce quotable evidence rather than opaque scores, and position Basanite as infrastructure that makes human judgement better, not the mechanism that replaces it.
A two-way mirror
The best hiring processes leave candidates with a clearer understanding of themselves. Every assessment strategy deployed by Basanite can be honestly explained to the candidate it is applied to.
Story so far
Six weeks. Three iterations. Live product.
From a shared frustration to a paid pilot.
Early 2026
Problem identified
The three of us were applying for graduate engineering roles and watching technical screens collapse: leaked question banks, take-homes that could be done by Cursor in ten minutes, and a hiring funnel that no longer measured anything real.
April 2026
MVP shipped
End-to-end working product: hirer dashboard, candidate portal, live 10–20 minute AI voice interview, dual reports grounded in candidate quotes. Built in week one.
May 2026
50 trial users
University of Manchester CS students, Manchester technology recruiters, and early hirers running mock interviews. Surveyed feedback drove iteration two.
May 2026
Stripe VC accelerator
Accepted into the Stripe internal accelerator alongside ongoing applications to YC and VFA26.
May 2026
First paid pilot
Verbal commitment from a seven-figure-revenue technology recruitment firm in Manchester. First contract worth ~£40k ARR.
2026 →
Iteration three
The AI Collaboration Workbench: a sandboxed VS Code environment where candidates ship a real ticket alongside the AI agent of their choice. The dimension no other interview measures.
The team
Built by people who felt the problem
Aditya Shah
CEO
Owns commercial, fundraising, customer discovery and positioning. Has lived the candidate side of the broken hiring funnel first-hand.
- ◆Data Analyst & Technology Modeller at Virgin Media O2 (13-month placement)
- ◆Previously founded an edtech startup
- ◆Final-year Computer Science, University of Manchester
Lynn Zhao
CPO
Owns product, primary research with hiring managers and occupational psychometricians, and the prompt architecture behind the interview agent.
- ◆BSc Artificial Intelligence, University of Manchester
- ◆AI Safety Fellowship at BlueDot Impact, prior internship at OpenAI Cambridge
- ◆UniHack 2025 Digital CleanUp, 1st place