Soslan D.

I help startups and product teams build practical computer vision solutions focused on video, gesture, and action recognition. My work is centered on delivering working MVPs rather than research prototypes. I design end-to-end pipelines that go from raw video data to reliable model inference, ready for demos, prototyping, or further product development. Core areas of expertise: • Gesture and hand recognition using keypoints (MediaPipe) • Action recognition from video (ST-GCN, transformer-based models) • Computer vision pipelines with PyTorch and OpenCV • Data preprocessing, model training, and inference optimization I focus on clear outcomes, clean code, and realistic timelines. If you need a reliable video-based CV solution for your product or demo, feel free to reach out.