Antech Studio · HR Tech · Recruitment
An AI co-pilot for recruiters — upload CVs in any format and get every candidate scored, ranked and summarized against the role, without taking the decision out of human hands.
2025
Year
3 weeks
Timeline
The Challenge
HR recruiters and talent teams need to screen large volumes of candidates quickly, but manual CV review is slow and inconsistent. Built as a hackathon project in Rwanda, the goal was to speed up screening without replacing the recruiter's judgment — an assistant, not a replacement.
What I Built
I built an AI-assisted recruiter and talent portal where HR teams upload CVs in multiple formats — PDF, Word, Excel and CSV. The system extracts candidate data automatically and uses Gemini AI to score and rank candidates against role requirements, surfacing each candidate's strengths and weaknesses with a full scoring metric and an AI-generated summary. Recruiters keep full control and can still screen manually at any point — the AI assists, it doesn't replace them. A queue-based async worker architecture runs extraction and AI scoring in the background, so the platform stays responsive regardless of upload volume.
- Multi-format CV upload — PDF, Word, Excel and CSV
- Automatic extraction of structured candidate data
- Gemini AI scoring and ranking against role requirements
- Full scoring metric with per-candidate strengths and weaknesses
- AI-generated summary for every candidate
- Manual screening — recruiters keep full control
- Async, queue-based processing that stays responsive at any volume
Built With
The hardest part was designing an async, queue-based pipeline — RabbitMQ plus a dedicated worker service — that could take CVs in many raw formats, extract structured candidate data reliably, and hand it off to the AI for scoring, all without blocking the main app or making recruiters wait on processing. Keeping extraction reliable across PDF, Word, Excel and CSV, then sequencing it into background AI scoring at any upload volume, was the core engineering problem.
Prototype
Working hackathon build — validated the full pipeline end to end; not yet commercially deployed
4 formats
PDF, Word, Excel and CSV parsed into structured candidate data
Built in 3 weeks at a hackathon in Rwanda.
Let's build the thing that gets you booked.
If you've got a problem worth solving — a system, a store, a platform — I'd like to hear about it. You talk directly to the person who'll build it.