Shanmukha Sreenivas

About Shanmukha Sreenivas

I am an AI / Machine Learning Engineer with 4+ years of experience designing, building…

Experience

1. Machine Learning Engineer

   DXC Technology

   Sep 2024 — Present · US

   Engineered a real-time fraud detection system in Python (scikit-learn, XGBoost, TensorFlow/PyTorch) on AWS SageMaker added temporal graph (TGN) and TCN sequence features, lifting AUC-PR by 22% and cutting FPR by 15%. • Developed scalable pipelines with Kafka, Apache Spark (PySpark/Spark SQL) on Databricks and Airflow for ETL, feature engineering, auto retraining; processed 10M+ transactions/day, reducing data latency by 50% and feature freshness under 1 min. • Integrated models as FastAPI + Docker microservices on Kubernetes (EKS); optimized inference via NVIDIA Triton serving ONNX Runtime + TensorRT (FP16/INT8, dynamic batching) to achieve p95 under 180 ms and approximately 3 times throughput. • Implemented MLOps with GitHub Actions, MLflow Model Registry, W&B, and Terraform; added drift monitoring (Evidently + Prometheus) and canary/shadow releases, shortening release cycle by 30%, reduced MTTR by 40%, maintained uptime ~99.9%. • Shipped a fraud-analyst copilot via LangChain+LlamaIndex (hybrid BM25 + dense over FAISS), LoRA-tuned LLM served by vLLM integrated SHAP evidence in dashboards, cutting triage time by 25%, and improved case resolution by12%. • Optimized cloud ML infrastructure on AWS (S3, Lambda, ECS, SageMaker) by leveraging auto-scaling and spot instances, reducing model training costs by 18% and inference costs by 12% while maintaining ~99% uptime and high availability.

Education

• Mahindra University

  Bachelor of Technology - BTech, Computer Science

• North Carolina State University

  Master of Science - MS, Computer Science