At Meta, I led end-to-end AI systems powering personalization at a massive scale fine-tuning LLMs (LLaMA 2, Mistral) for multimodal summarization, building real-time recommendation engines with BERT & reinforcement learning, and engineering petabyte-scale data pipelines using PyTorch, Spark, and TensorFlow. On the infrastructure side, I automated CI/CD workflows via Kubernetes, FBLearner, and PyTorch Serve — cutting deployment cycles from weeks to days while maintaining 99.9% uptime. The result? Measurable lifts in ad relevance, feature extraction efficiency, and user engagement all while reducing compute costs through model pruning and quantization. From model to production, I owned the full stack.

At Freshworks, I built the ML backbone that made SaaS smarter designing and deploying end-to-end models using XGBoost and deep learning for lead scoring, sentiment analysis, and automated ticket classification, directly boosting agent productivity and automation coverage. I architected the full ML infrastructure on AWS (SageMaker, Lambda, ECS), built large-scale feature engineering pipelines with Python and PySpark, and shipped production-ready APIs via FastAPI and Docker cutting inference latency and enabling clean microservices integration. Beyond building, I put in place monitoring and drift-detection frameworks to keep models reliable long after deployment. All of it delivered within Agile teams, aligned tightly to product goals and business outcomes. End-to-end ownership, zero hand-waving.