Shivansh Chakrawarti

Reduced deployment time by 50% through containerization of Spring Boot and ML workloads, automating CI/CD pipelines with Jenkins on AWS ECS and GCP GKE.

Engineered dynamic autoscaling solutions, reducing cloud costs by $40,000 by leveraging KEDA with GPU metrics, Kafka queue optimization, and enhanced observability via Prometheus.

Optimized Reactive Spring Boot ingestion, cutting database calls by 90% through re-architecting and implementing strategic batch processing to prevent connection overload during peak loads.

Developed a resilient Change Data Capture (CDC) pipeline, achieving sub-100 ms latency at 10K+ events/sec by streaming real-time updates from PostgreSQL and MongoDB into Elasticsearch using Debezium and Apache Kafka.

Developed scalable Spring Boot microservices with robust RESTful APIs, integrating a centralized Eureka server to streamline business operations through seamless service discovery.

Automated the end-to-end churn-prediction ML lifecycle using MLflow and Databricks, integrating GitHub Actions for CI/CD.

Authored a comprehensive MLOps reference architecture, significantly boosting reliability and cutting cloud costs for scalable model training.

Conducted research on optimizing fixed-time convergence of gradient flows, co-authoring an internal publication.

Benchmarked leading deep learning models including ResNet, VGG, and Transformers in PyTorch, achieving 95% accuracy on CIFAR/MNIST datasets.