Architected enterprise-grade AI solutions for Microsoft Copilot using Azure OpenAI, GPT-4, and RAG with Azure AI Search over 15M+ documents. Built ingestion and indexing pipelines to reduce indexing time, and implemented hybrid retrieval (vector/semantic, BM25, metadata filters, cross-encoder reranking) to improve retrieval precision and reduce hallucinations. Developed multi-agent Copilot workflows with LangGraph and function calling for summarization and knowledge discovery, and implemented LLM evaluation and Responsible AI guardrails (PII detection, prompt-injection protection, moderation, RBAC, audit logging). Designed scalable inference infrastructure using FastAPI, Docker, Kubernetes, Azure ML, and monitoring, reducing latency while supporting high request volumes. Partnered cross-functionally to deploy production-ready capabilities and improve adoption and user productivity.

AI/ML Engineer at Netflix (AI-Powered Content Personalization & Recommendation Platform)

February 1, 2019 - May 1, 2023

Architected and deployed deep learning recommendation systems combining collaborative filtering, content-based retrieval, and transformer-based ranking to improve recommendation relevance and viewer engagement. Built large-scale candidate retrieval and ranking pipelines leveraging behavioral signals, features, and content embeddings to increase CTR and watch time. Developed feature engineering pipelines using Python, Spark, Kafka, and Airflow to process billions of events with reduced latency and enabled near real-time personalization. Implemented online A/B experimentation for engagement and retention metrics, and engineered low-latency model serving and MLOps pipelines for automated training, deployment, monitoring, drift detection, and continuous retraining.