AI Engineer Internship (Computer Vision)

Responsibilities • Develop Computer Vision solutions using Python and OpenCV. • Work with YOLO / Ultralytics and other deep learning-based Computer Vision models. • Prepare, clean, organize, and annotate image/video datasets. • Train, fine-tune, validate, and evaluate Computer Vision models. • Work with object detection, classification, tracking, recognition, OCR, and video analytics. • Build image and video processing pipelines for real-world applications. • Work with CCTV footage, RTSP streams, and real-time video processing. • Research and experiment with different Computer Vision techniques, models, papers, and open-source implementations. • Develop Proofs of Concept (POCs) to test new ideas and approaches. Model Evaluation & Optimization • Evaluate models using metrics such as: • Precision • Recall • mAP • F1 Score • IoU • Perform error analysis and identify reasons behind false positives and false negatives. • Improve model performance through better datasets, training strategies, parameters, and model selection. • Analyze real-world model performance rather than relying only on training metrics. • Experiment with different image sizes, augmentations, models, and inference approaches. AI Integration & Deployment • Integrate trained AI models into Python-based applications and APIs. • Work with frameworks such as Flask or FastAPI where required. • Develop and debug AI inference pipelines. • Work with databases and backend systems when required. • Assist in deploying Computer Vision solutions in real-world environments. • Optimize AI pipelines for performance, reliability, and scalability. • Work with technologies such as Docker, Linux, cloud platforms, CUDA, ONNX, or TensorRT when applicable. Research & Problem Solving • Research technical problems independently before seeking help. • Read documentation, papers, GitHub repositories, and technical resources to understand possible solutions. • Compare different approaches and explain why a particular solution was selected. • Build small experiments and POCs to validate ideas. • Troubleshoot model and pipeline issues systematically. • Document experiments, findings, and technical decisions.