Lead Machine Learning Engineer at Apple

Led development of the Apple ML Platform using Kubernetes, Apache Kafka, Apache Flink, Spark Streaming, Redis, Cassandra, DynamoDB, gRPC, and PyTorch to support multi-million user workloads with 99.99% service availability across distributed ML infrastructure. Architected scalable ML services for personalization and recommender systems using Kubernetes, Redis, and gRPC microservices, reducing model inference latency by 45% and increasing throughput by 3.5x. Designed and deployed streaming ML pipelines with Kafka, Flink, Spark Streaming, Kubernetes, and Snowflake to process billions of user events daily for near real-time feature engineering and model updates. Spearheaded embedding systems and ranking models using PyTorch, TensorFlow, and Feature Engineering pipelines, boosting recommendation engagement by 22%. Built enterprise-scale ML Ops infrastructure with Jenkins, GitLab CI/CD, Terraform, CloudFormation, Docker, and Kubernetes, enabling multiple daily deployments. Led architecture

Senior Data & Machine Learning Architect at Amazon Web Services (AWS)

Led architecture for Identity Graph Platform using Graph Neural Networks, Apache Spark, Redis, Cassandra, DynamoDB, and gRPC to support cross-device identity resolution, achieving 90% precision and recall and reducing processing time from 7 days to under 6 hours. Built recommendation systems for Sports News and CNN News using Deep Learning, embedded models, user behavior analytics, and A/B testing, increasing user engagement metrics by 18%. Designed and deployed real-time forecasting systems for business KPIs, analytics, and executive dashboards, contributing to a 30% increase in signed enterprise deals. Engineered scalable distributed data pipelines and streaming analytics handling billions of customer interaction events with Apache Spark, Apache Kafka, Hadoop, and cloud-native infrastructure. Implemented low-latency inferences services, automated ML retraining workflows, and scalable real-time prediction systems that improved model freshness, prediction accuracy, and production relia