Led the development of Transfer + Buy and Sell + Transfer, enabling customers to move funds and invest or liquidate holdings in a single streamlined flow. Owned these projects end-to-end from RFC design and system architecture to implementation and launch.

Also contributed to control account relationships, net yield updates, several Arc Capital Markets initiatives, and improvements to borrower onboarding flows and the lender portal.

Worked on Optimizing the Deep Neural Network models for Traffic Flow Prediction in Intelligent Transport Systems. Applied TinyML optimization techniques (model pruning, quantization, transformer distillation) for low-latency inference. Attained 25% reduction in memory size and 15% faster inference time for real-time deployment in embedded traffic sensors. Published research findings in IEEE VT Conference, highlighting optimizations for Intelligent Transport Systems.

Implemented an advanced cache architecture and deployed an efficient pipeline for QAS Deep Neural Network models achieving 60% reduction in CPU utilization and cost by optimizing memory access patterns and avoiding redundant computation. Engineered real-time cache update and deletion logic, significantly boosting inference accuracy and performance.

Developed an application as a part of the Indo-Japan joint research laboratory project, focusing on guiding farmers to effectively tackle crop diseases. Implemented authentication, navigation, and integrated Cytoscape graph for efficient information access and strategy exploration.