• Develop and deploy vision-based perception algorithms for autonomous inspection robots using deep learning and machine vision techniques, including CNN- and ViT-based architectures • Design and implement supervised, self-supervised, and unsupervised learning methods for multi-modal data (RGB, LiDAR, thermal, etc.) • Build and optimize models for: Object detection and tracking (2D & 3D) • Anomaly detection • LiDAR-based semantic segmentation and 3D point cloud understanding • Optimize models for on-device inference (quantization, pruning, distillation) and deploy them on edge hardware (NVIDIA Jetson Xavier/Orin/Thor, etc.) • Integrate perception pipelines into robotic systems via ROS1/ROS2 and collaborate closely with software and robotics teams • Support MLOps workflows, ensuring reproducible training, evaluation, and CI/CD for AI models • Support and maintain cloud-based AI pipelines on AWS (SageMaker, Bedrock, Lambda, S3) for scalable training, inference, and model lifecycle management • Contribute to research on multi-modal and 3D representations (Gaussian Splatting, NeRF, OpenCLIP-based embeddings) • Participate in continuous improvement of our AI infrastructure for mission planning, semantic mapping, and robot autonomy