Computer Vision Engineer Internship

About Us:

• We are building AI managers for physical businesses which would automate their daily tasks and would help reduce human labour and infrastructure cost accordingly. We are creating a store super manager which would keep an eye (literally, through cctv cameras) on all the existing processes and take data from other source of truths like HR, POS and Inventory data.

Responsibilities:

• Build, test, and deploy production-grade computer vision use cases for physical businesses
• Design and maintain training and inference pipelines for vision models
• Work with large-scale CCTV video data (long-duration, multi-camera feeds)
• Optimize models for accuracy, latency, and reliability in real-world conditions
• Integrate vision outputs with backend systems such as HR, POS, and inventory data
• Perform data validation, error analysis, and continuous model improvement
• Own models end-to-end: from data ingestion → training → deployment → monitoring
• Collaborate with product and operations teams to convert business problems into ML solutions

Requirements:

• Strong fundamentals in computer vision and deep learning
• Hands-on experience with Python and at least one deep learning framework (PyTorch / TensorFlow)
• Experience building training pipelines (data loading, augmentation, training, evaluation)
• Familiarity with object detection, tracking, activity recognition, or face recognition
• Experience deploying models to production environments (edge or cloud)
• Understanding of Linux environments, Docker, and basic DevOps concepts
• Ability to debug model failures using real-world data (not curated datasets)
• Comfortable working in a fast-moving startup environment

Good to Have:

• Experience with real-time video processing and multi-camera systems
• Prior work on retail, surveillance, or industrial vision systems
• Knowledge of model optimization (TensorRT, ONNX, quantization, pruning)
• Experience working with large unstructured datasets
• Familiarity with AWS / GCP / on-prem deployments