Zenithr
AI Engineer Intern
June 2026 to Present
Hybrid
- Built an LLM candidate-sourcing platform turning hiring briefs into ranked, evidence-linked shortlists: 350 candidates over 50 live searches, 98% completion, 18s median latency, $0.008 per search.
- Shipped an LLM extraction and ranking pipeline for web profiles and PDFs at 98.3% field-level accuracy, 0.99 F1, and 100% schema-valid output across a 345-field ground-truth set.
- Architected a Celery and Redis crawling backend with HTTP-first fetching and Playwright fallback, sustaining 20 concurrent jobs at 1,780 pages/min with 3.0x worker scaling and 1,116 tests.
