StudentVenture

The Challenge

StudentVenture is a dual-sided recruitment marketplace connecting early-talent students with startups and scale-ups. The platform had over 10,000 student profiles. The problem: no way to match them to relevant employers at scale. Traditional keyword search missed 80% of relevant fits. Candidates were invisible to recruiters unless their CV contained the exact right words.

The Approach

We designed and built a semantic search system using OpenAI embeddings, pgvector on PostgreSQL, and a custom LLM-based grading layer. Student profiles are embedded at creation and on update. When an employer posts a job brief (created conversationally by an AI chatbot), the system performs approximate nearest-neighbor search over all 10,000+ embeddings, then grades the top candidates with a fine-tuned LLM scorer. The result is ranked, relevant candidates in under one second.

The Result

The platform shipped to production and closed its funding round. Sub-second semantic matching is live. The founders had a clear reaction.

It feels like magic.

Founder, StudentVenture