AI Job Matching MVP Cuts Time-to-Hire to 38 Hrs
The job matching platform MVP is an AI recruiting product built for Resumind in 4 weeks, with semantic resume search, role-fit scoring and automatic outreach drafts. Under the hood, an e5-large embedding pipeline and a domain re-ranker feed hybrid search over Postgres pgvector, multi-step matching agents score by role rather than keywords, and a Next.js recruiter dashboard drafts personalized outreach in the recruiter's voice.
A purpose-built embedding pipeline using e5-large + a domain re-ranker, hybrid search over Postgres pgvector, multi-step matching agents that score by role rather than keywords, and a Next.js recruiter dashboard that drafts personalized outreach in the recruiter's voice.
Book a free 30-minute callHR Tech
Industry
4 Weeks
Timeline
Project Overview
1
What We Built
- A purpose-built embedding pipeline using e5-large + a domain re-ranker, hybrid search over Postgres pgvector, multi-step matching agents that score by role rather than keywords, and a Next.js recruiter dashboard that drafts personalized outreach in the recruiter's voice.
2
The Challenge
- Resumind's two co-founders raised $1.1M to compete in a market dominated by ATS giants. They needed a working AI job-matching MVP, semantic resume search, role-fit scoring, automatic outreach drafts, before their seed round LOIs expired in five weeks. Every off-the-shelf RAG tutorial broke at 50k resumes.
3
Our Solution
- Semantic + keyword hybrid search over 1M+ resume embeddings
- Role-fit scoring with explainable feature breakdowns recruiters can edit
- Voice-cloned outreach drafts respecting each recruiter's tone library
- Eval suite tracking precision@k on labelled hire/no-hire pairs
- ATS write-back via Greenhouse, Lever, and Ashby APIs