Projects

Developed a computer vision system using fine-tuned ResNet architectures to assist radiologists in detecting anomalies in X-rays, achieving 94% accuracy and reducing average review time by 30%.

Created an open-source toolkit to automatically refine and optimize prompts across 15+ languages using reinforcement learning, reducing token usage by 20% while improving multilingual inference accuracy by 15%.

Simulated an urban traffic management system evaluating PPO and DDPG algorithms to dynamically adjust traffic light timings, decreasing average waiting times at intersections by 25% in simulated environments.

Built a microservice for semantic codebase search leveraging AST parsing and vector embeddings of code snippets, improving search relevance score by 40% compared to traditional grep-based searches.