Kushagra Bhargava

I build production ML pipelines end to end—modeling, deployment, monitoring, and iteration—so teams get measurable impact quickly. Most recently, as an ML & Analytics Intern, I improved churn model performance by 18% (ROC-AUC 0.91) across 100K+ records and reduced stakeholder time-to-insight from days to under 30 seconds by shipping a 4-tier FastAPI scoring service with automated Airflow ingestion and batch inference.

I’m especially energized by reliability and real-world deployment: in SentinelML, I implemented KS-test/chi-squared drift detection across 7 drift types and triggered Airflow-orchestrated retraining with zero manual intervention, plus a 3-gate MLflow promotion system for safer releases. I also deliver pragmatic forecasting and retrieval solutions—MediFlow achieved RMSE 1.28 for 7-day inflow forecasting and reduced ICU overcapacity projections by 18% using PuLP constraints, while EnterpriseRAG boosted Precision@5 from 0.62 to 0.81 and reached 2s query latency with metadata-filtered RAG. Alongside delivery, I share what I learn through Substack and contribute to open source (Giskard AI), backed by an Oracle Data Science Professional certification.