VLM Internship

What You'll Work On

• Develop 2D computer vision pipelines for object detection, classification, segmentation, tracking, and image analysis.
• Work with stereo, RGB-D, and depth cameras for 3D perception and scene understanding.
• Process 3D point clouds using Open3D and Python-based tools.
• Work on camera calibration, depth processing, coordinate transformations, and sensor-to-robot calibration.
• Develop perception pipelines using OpenCV, NumPy, SciPy, PyTorch/TorchVision.
• Experiment with Vision-Language Models (VLMs) for image understanding, visual reasoning, object identification, and task interpretation.
• Explore Vision-Language-Action Models (VLAMs), foundation models, and embodied AI for robotic applications.
• Build datasets, prepare annotations, evaluate models, and perform inference/benchmarking.
• Integrate vision and AI pipelines with robotic applications and real-world sensors.

Must Have

• Bachelor's/Master's degree or final-year student in Computer Science, AI/ML, Robotics, Mechatronics, Electronics, or related field.
• Strong fundamentals in Python.
• Basic-to-good understanding of Computer Vision and image processing.
• Hands-on exposure to OpenCV, NumPy, and PyTorch/TorchVision.
• Understanding of basic 2D/3D geometry, depth, point clouds, coordinate frames, and transformations.
• Familiarity with Linux/Ubuntu and Git.

Good to Have

• Open3D and 3D point-cloud processing.
• RGB-D/stereo cameras such as Orbbec, RealSense, or equivalent.
• YOLO, Grounding DINO, SAM/SAM2, or equivalent models.
• VLMs such as Qwen-VL or similar multimodal models.
• ROS2, robotics simulation, Docker, CUDA, or NVIDIA platforms.
• Camera/hand-eye calibration or exposure to robotics and embodied AI.