Skip to content
Shivam Mishra
Selected work

AI-Assisted Candidate Screening

A recruiter-facing screening assistant that combines embedding search and LLM ranking to match candidates to job descriptions, generate technical questions, and produce structured candidate reports.

Capability flow

  1. 01Résumés + JD
  2. 02Parsing
  3. 03Embedding search
  4. 04LLM ranking
  5. 05Questions + report

The problem

Manual résumé screening is slow and inconsistent at volume. Recruiters spend much of their time reading and comparing profiles before they ever talk to a candidate.

What the system does

  • Parses résumés into structured candidate information.
  • Matches candidates to job descriptions using embeddings and vector search, rather than keyword overlap alone.
  • Uses an LLM to rank and reason over the shortlist.
  • Generates technical screening questions in the context of both the role and the candidate.
  • Produces structured, recruiter-ready candidate reports.

Approach

The assistant is built as a retrieval-augmented generation (RAG) pipeline. Embedding search narrows the candidate pool against the job description, and LLM ranking, question generation, and reporting work from that retrieved context.

My role

I built the RAG-based screening assistant, combining embedding search and LLM ranking, and worked closely with enterprise clients to deploy AI-driven screening in their hiring workflows.