Ayush T.

I build and deploy computer vision and multimodal AI systems for factories, camera networks, and robotics—from data and models to edge inference and operator tools. I have 6+ years of experience in computer vision, robotics, and applied AI, including three years at a robotics company. As the founder of AI2SEE, I lead technical delivery and bring in specialist team members when the scope calls for them. My work covers the complete pipeline: camera and lighting setup, data preparation, model training, detection and tracking, optimization, integration, and deployment. I account for the conditions the system actually operates in—changing lighting, noisy inputs, latency requirements, and hardware limits. Selected projects include industrial parts counting with an operator interface; edge ANPR and OCR; multi-camera video analytics using DeepStream and TensorRT; and workout video analysis with exercise recognition and repetition timing. In my previous robotics role, I worked on stereo depth and floor-obstacle mapping for production robots. I also work with vision-language models and LLMs for multimodal understanding, structured extraction, and tool-assisted workflows. I select the approach around the task, deployment constraints, and evaluation criteria. I can help you build a new system, improve an existing model or inference pipeline, or maintain and extend a deployed solution. We agree on acceptance criteria first, review progress on your real inputs, and define deployment, handover, and ongoing support. Core tools: Python, PyTorch, OpenCV, YOLO, ONNX, TensorRT, NVIDIA Jetson, DeepStream, FastAPI, ROS, and C++. Send me an example input, the output you need, and any hardware constraints. I’ll outline an approach, a useful first milestone, and the main technical risks.