LLM + MLOps
KIRU
AI-powered HR platform for bulk resume analysis
- React
- Redux
- Django
- LLMs
- Neural Networks
- MLOps
Challenge
Recruiters needed to score and rank thousands of resumes against job descriptions in minutes, not days, with explainable matching.
What I built
- Built Django backend orchestrating LLM inference + classical neural ranking
- Designed React/Redux frontend for bulk uploads and side-by-side comparisons
- Set up MLOps pipeline for reproducible model training and deployment
Key decisions
- Used LLM embeddings for semantic match + a smaller neural ranker for explainability
- Versioned models through the MLOps pipeline so HR teams could roll back regressions
Outcomes
- Reduced screening time from days to minutes
- Reproducible deployments via standardized MLOps workflow