My Projects
Here are some of the projects I've been working on. You can find more on my GitHub.
Eytrix
- Status:
- Acquired by Meta in August 2025
- Website:
- eytrix.com
Led the design and implementation of an AI-driven video surveillance platform, specializing in computer vision, distributed inference, and scalable backend systems. Architected real-time detection models for events and anomalies using multi-modal inputs and reference-based similarity matching. Built a continual learning pipeline that enabled organizations to upload reference images and automatically update detection models. To support government and enterprise environments, developed a distributed inference system optimized for cloud, on-prem, and edge deployments, delivering low-latency performance without requiring new camera equipment. On the systems side, designed high-throughput, multi-camera video processing pipelines for event filtering and analytics, supported by a microservices architecture with RBAC, audit logging, SSO integration, and automated reporting enriched with model insights and snapshots. Also led development of enterprise features such as customizable detection policies, mobile and SMS alerts, and organization-specific model configurations, while overseeing privacy, compliance, and security requirements for government deployments.
Adaptive Streaming of Multimedia Content
- Associated with:
- Netflix
Developed a computer-implemented method for adaptive streaming of multimedia content, focusing on dynamically adjusting audio quality while maintaining a specified minimum video quality. The system ensures seamless multimedia streaming by prioritizing video quality and dynamically adjusting the audio quality in real-time based on network conditions. This approach optimizes the user experience by providing consistent video quality while adapting audio quality to available bandwidth.
Content-Based Split Tunneling for Network Traffic
- Associated with:
- Netflix
Developed a system for split tunneling based on content type to selectively exclude certain network traffic from a VPN tunnel. The solution involved monitoring session traffic, identifying if the traffic corresponds to a specific content type, and dynamically redirecting it based on a predefined policy. This approach optimizes network performance by routing traffic based on its type, ensuring more efficient use of the tunnel and improving overall system efficiency.