Designed and implemented the first-of-its-kind reinforcement-learning agent to intelligently navigate complex IoT device environments—reducing exploration latency by 35% and boosting malware detection coverage by 25% over a brute-force search baseline.

Implemented a heterogeneous graph neural network combined with autoencoders to model inter-device relationships and inform RL agent policy, aiming to enhance policy convergence and improve malware detection accuracy.

Designed the operational framework for a new founder hub, bridging the gap between customer discovery and technical building.

Solely developed the organization's operating platform (Next.js/TypeScript), implementing secure RBAC pipelines to automate student and alumni management.

Engineered and deployed a customer-facing, scalable video memory framework for AR apps, enabling long-term spatial and contextual awareness and optimizing throughput for speed and scale.

Designed and launched a Python SDK for the Mavi platform, driving end-to-end developer workflows around video analysis; published to PyPI at pypi.org/project/pymavi

Collaborated in a cross-functional, iterative team of engineers and researchers to build end-to-end real-time SLAM and novel view-synthesis pipelines, improving scene reconstruction accuracy by 25% while ensuring safety and reliability in deployment.

Implemented performance optimizations in C++ and Python to scale mapping algorithms for autonomous navigation, reducing processing latency