Vijay Dubey

I've built and deployed AI-driven perception systems for robotic manipulation at DRDO-CAIR, enabling adaptive grasp generation across 21 object types. My YOLOv8, PyTorch, and OpenCV stereo-vision pipelines achieved 94.8% mAP@0.5 while reducing perception-to-grasp latency to 165 ms.

At Futuristic Bots, I developed camera-based robotic perception for pallet detection and localization, supporting more than 2,000 autonomous pickup operations. I implemented AprilTag and YOLO-based vision systems that improved pickup success and reduced false detections under variable lighting.

I also build Generative AI applications, including RAG pipelines with LangChain, Azure OpenAI, Vertex AI, FAISS, MLflow, and Ollama. I bring hands-on experience deploying real-time models with Docker, Kubernetes, ROS 2, and production robotics workflows.