LLM / RAG · BigCJobs
Resume Intelligence Platform
AI-powered talent repository with RAG search over ~20,000 resumes, built on an Agentic Harness system
FID ·
Harsha's platform read ~20,000 resumes with OCR, embeddings and vector search, so HR can ask in plain English. I asked for my own resume. Empty.
- ~20k
- resumes processed
- 8–10 h/wk
- saved per HR employee
Problem
HR teams spent hours hunting candidate data across resumes, assessments and trackers.
What Harsha built
Built an Agentic Harness system for document ingestion, native text extraction with OCR fallback, structured field extraction, duplicate detection, and RAG search over ~20,000 resumes.
Outcome
~20,000 resumes processed into a searchable Candidate Master; an estimated 8–10 hours per week saved per HR employee.
Stack
- Python
- OCR
- RAG
- Embeddings
- Vector Search
- PostgreSQL
- LLM extraction
Next projectVinayaga Timbers ERP