AI/ML Computer Vision Engineer at Vinfotech (2025-03 – Present) • Facial recognition-based attendance system: Integrated YOLO-based face detection with MediaPipe to score face quality (yaw, pitch, roll, landmark spread) and implemented frame filtering to select the best quality faces.
- Optimized pipeline by encoding filtered faces with a deep learning model and sending only unknowns to AWS Rekognition, reducing API calls by 40% as cost.
- Quantized model to ONNX for Jetson Orin Nano and optimized frame processing via multithreading and larger RTSP buffer to prevent frame drops.
- RTSP Video Streaming Pipeline: Developed a video streaming system to ingest RTSP streams over TCP from NVR cameras, re-encoded using FFmpeg for optimized playback and delivered via a MediaMTX server for seamless live monitoring and video analytics.
Computer Vision
Engineer at Camncloud (2023-11 – 2025-02) • QR/Barcode scanner on a moving conveyor belt: Trained YOLO model for real-time QR/barcode detection on fast-moving objects.
- Implemented an image processing pipeline with techniques like Gaussian blur, adaptive thresholding, and others to improve detection accuracy by 16%.
- GeoSyncVision: Satellite Data and Video Frame Mapping: Engineered a dual-camera setup to capture synchronized footage and construct seamless panoramic views.
- Developed a system to map geographic coordinates (latitude and longitude) from satellite data to individual video frames by timestamp alignment.
- Overlaid geographic data on each frame and converted them into a synchronized video.
- BoundaryGuard: Automated Trespass Detection: Developed a trespassing detection system using YOLOv8 with line-crossing logic and polygonal ROIs via OpenCV's pointPolygonTest for accurate, real-time violation alerts.
- Reducing system load by 20% with an adaptive frame skipping mechanism that dynamically resumes processing.
Web Application Developer
Intern at SUREBOT (2022-03 – 2022-05) • Developed REST APIs for HRMS application: Utilized Django REST framework and SQL for efficient data handling and backend processes.
