Computer Vision

Responsibilities :

Assist in building and testing CV models for detection, tracking, classification tasks and latest foundation and VLM models. Prepare and annotate image/video datasets, support data ingestion and cleaning pipelines Contribute to writing and debugging training scripts, model loaders, and preprocessing functions Run evaluation jobs and generate performance reports using tools like TensorBoard or custom scripts Support error analysis by identifying model weaknesses across edge cases Collaborate with senior engineers on integrating models into scalable inference pipelines Help visualize model outputs, draw bounding boxes, heatmaps, or segmentation masks for explainability Document experiments and code for reproducibility and knowledge sharing Used OpenCV, MediaPipe, and scikit-image for preprocessing, motion analysis, and visual overlays Integrated DL models with post-processing logic (e.g., NMS, temporal smoothing, event triggering) Ensured low-latency inference by profiling and tuning frame-wise preprocessing Supported integration of RTSP video feeds and video decoders in test pipelines point-cloud ingestion and processing (Open3D/PCL), calibration/registration (ICP/FGR), 3D detection/segmentation with sparse CNNs/PointNet, and RGB-LiDAR-IMU fusion.

Additional Responsibilities:

Basic understanding of synthetic data or 3D asset usage in training pipelines Familiarity with Git, Linux command line, and Jupyter Notebooks Interest in building a career in computer vision and AI for real-world automation Bachelor's degree in Computer Science, Data Science, or a related technical field 1-3 years of experience or strong internship/projects in computer vision or ML model development

Technical and Professional Requirements:

Good Python skills and working knowledge of PyTorch or TensorFlow Familiarity with image processing libraries (e.g., OpenCV, PIL) and dataset tools (e.g., COCO format, YOLO datasets) Eagerness to learn, take feedback, and contribute in collaborative development environments Exposure to object detection/tracking projects (academic, hackathons, or prior work)