Timeline: Oct 2025
Led a hackathon team that fused NASA Earth observation data, reinforcement learning, and geospatial analytics to deliver personalised rainfall forecasts for event planners. The live demo showcases location-aware precipitation probability maps, confidence metrics, and AI-driven insights powered by Random Forest ensembles trained on IMERG rainfall archives.
The application allows users to specify dates and locations, returning interpretable summaries of the likelihood of “very wet” or other adverse weather conditions. It integrates NASA POWER datasets, Giovanni workflows, and custom reinforcement learning pipelines that optimise query planning to reduce latency. Planned extensions include mobile notifications, POWER-based thermal indicators, and advanced AI models for hyper-local forecasts.
Timeline: Nov 2024 – May 2025
Built an Azure ML pipeline that automates real-time inference workflows using gradient boosted trees, maximizing AUC/ROC performance for production deployments.
Timeline: Jun 2022 – Aug 2022
Designed a DQN and PPO driven controller for IoT infrastructure, reducing power consumption by 20% against baseline methods while preserving service responsiveness.
Timeline: Jan 2025
Achieved 92% accuracy in real-time drowsiness detection through facial landmark analysis with OpenCV, Dlib, and machine learning classifiers to enhance driver safety systems.
Timeline: Jan 2022 – Apr 2022
Deployed TensorFlow and PyTorch pipelines for autonomous vehicle perception, surpassing 90% detection and 92% accuracy while automating 80% of vision and NLP workflows.
Timeline: Jun 2021 – Jul 2021
Integrated multispectral MFNet data to reach 95% fusion accuracy, enabling reliable low-visibility object detection for perception systems.
