AI Job Matching
Resumes and vacancies are assembled in a dialogue, and matching runs on vector search
What it gave the business
A person describes themselves in words — and finishes, instead of abandoning a form at the fifth field. Matching gets sharper too: it searches by meaning, not word overlap.
The problem
A twenty-field form kills conversion, and keyword search does not know that babysitting and childcare are the same thing. Both had to go.
My role
The app, a database with vector search, the conversational onboarding and matching, plus multilingual support with auto-translation.
What was built
- The employer describes the job by voice or text; a dialogue clarifies details and assembles a structured description
- The candidate builds a resume the same way — a five-step dialogue instead of a form
- Matching runs on vector search by meaning rather than word overlap
- Chat with message translation, reviews, complaints and saved listings
- Three interface languages with automatic translation of user content
The engineering core
The best example here of AI as an input method: a conversation goes in, a structured database record comes out, and vector search runs on top of it. The core flow is closed end to end — sign-up, resume, matching, chat, review.
Stack
Tell me about your project
I will tell you how long it takes and what it costs before any work starts. If it is not my kind of task, I will say so right away and point you elsewhere if I can.
Message on Telegram