TalentLens — AI Hiring & Talent Intelligence

A mid-sized tech firm was drowning in 800+ applications per role. Their HR team spent 3 weeks per hire just reading CVs. TalentLens automated the entire top-of-funnel: ingesting résumés in any format, embedding them alongside job descriptions, scoring semantic fit, flagging red lines, and generating interview briefs — so recruiters only touched the top 10 candidates.

• 1

  Résumés arrive in PDF, DOCX, LinkedIn exports, and plain text — each with wildly different layouts and no standardisation.

• 2

  Generic keyword matching surfaces people who list skills without evidence — the client needed proof-based scoring, not keyword bingo.

• 3

  Automated screening triggers legal scrutiny — the system had to be explainable, auditable, and bias-tested before deployment.

• 4

  Calendar coordination across 6 timezones was a manual bottleneck even after screening was finished.

We built a multi-stage pipeline: a document parser normalises résumés into structured JSON regardless of format. A fine-tuned embedding model (built on `text-embedding-3-large`) projects résumés and job descriptions into the same latent space — candidates are ranked by cosine similarity, not keywords. A rubric layer lets hiring managers weight criteria (years of experience, specific tech, leadership signals) without writing code. Interview briefs are auto-generated by GPT-4o, citing the exact résumé sentences that drove the score. Finally, a Calendly-like scheduling micro-service sends personalised invites and collects availability with zero recruiter touch.