Case study · Recruitment and HR-tech

An assessment platform that writes its own tests

Recruiters needed to assess candidates consistently and at volume. Building each test by hand does not scale, and buying a fixed test library means assessing every role against the same questions whether they fit or not.

AI-generated tests and insights Dual candidate and recruiter portals Audit trail on approvals

Consistency is the hard part, not volume

Assessing more candidates is a throughput problem and it has obvious answers. Assessing them consistently is harder, because two recruiters marking the same interview will not agree, and neither will the same recruiter on a Friday afternoon.

What we built

Dual portals, one for candidates and one for recruiters. Tests are generated rather than hand-written, across English, Excel, programming, logical reasoning and psychometrics. Candidate insights cover skills gaps and cultural fit. A multi-stage approval workflow sits over it with an audit trail, so a hiring decision can be reconstructed later, which is the part HR actually needs when a decision is questioned.

Where the AI runs, plainly

Test generation and candidate insights run on GPT-4o, with Google Cloud TTS for the spoken components. This platform is not self-hosted. For generated assessment content the model quality is the product, and we would not trade it for hosting on this build.

Stack

React and TypeScript, Node and Express, PostgreSQL. GPT-4o for generation and insights, Google Cloud TTS for audio.

Want the same thing looked at in your business?

We map the process, find the part worth automating first, and give you a written scope with the phases separated. No obligation, and you keep the plan.