• Continued development and deployment of system to predict and prevent failures in Bar and Rod Mill of Bhilai Steel Plant using Time Series Anomaly Detection.
• Addressed concept drift and multimodal data distribution with continual learning and custom data normalization approach.
• I led the development and deployment of custom PLC signal logic in Python with minimum latency to detect event-based failures, mechanical issues, and sensor faults.
• I manage and maintain the system with added automation for stability.
• Stack: Scikit-Learn, PyTorch, Metaflow (MLOps Framework), PostgreSQL, InfluxDB, Redis, Grafana

• Worked on federated-split learning applications for bio-medical 3D Segmentation tasks.
• Stack: PyTorch, MONAI, nnUNet

Co-Author: Beyond Federated Learning for IoT: Efficient Split Learning With Caching and Model Customization [IEEE Xplore]

Applied split-learning on biomedical datasets for 3D segmentation with emphasis on communication and computation overhead comparison, model-splitting methods and trade-offs. My experiments improved metrics by 30% while reducing computation by 29x