**About Agrex.ai**
Agrex.ai is an innovative video analytics company dedicated to transforming conventional CCTV infrastructure into intelligent, real-time monitoring systems. Our technology serves a diverse range of industries, including retail, manufacturing, banking, logistics, and education. By leveraging advanced people and vehicle analytics, SOP/compliance monitoring, and operational insights, our platform operates seamlessly on both edge devices—such as NVIDIA Jetson—and cloud environments, delivering scalable, actionable intelligence to our clients.
**Position Overview: Computer Vision Intern**
We are seeking a motivated and skilled Computer Vision Intern eager to gain practical experience and contribute meaningfully to cutting-edge video analytics solutions. This internship offers a unique opportunity to work closely with our Product and Engineering teams, gaining hands-on exposure to model training, optimization, video pipeline development, and deployment on edge and GPU platforms. The ideal candidate will be passionate about computer vision applications and driven to solve real-world challenges in video analytics.
**Key Responsibilities**
– Assist in the design, training, and evaluation of advanced detection, segmentation, and tracking models, utilizing frameworks such as YOLO, RTDETR, Mask models, ByteTrack, and OC-SORT.
– Support the integration of machine learning models into NVIDIA DeepStream and GStreamer pipelines, including exporting models to ONNX and TensorRT formats, with a focus on optimizing for computational efficiency and real-time performance.
– Contribute to the development and enhancement of core features including people and vehicle counting, dwell time and queue analysis, intrusion detection, personal protective equipment (PPE) compliance verification, and simple re-identification (re-ID) tasks.
– Work extensively with real-time video streams in various formats (RTSP, H.264, H.265), handling challenges such as frame sampling, jitter management, and stream reconnection to ensure robustness in production environments.
– Write clean, maintainable Python code, and explore performance-critical sections using C++. Participate in unit testing, benchmarking, and code reviews to maintain software quality.
– Support data pipeline operations including dataset curation, establishing labeling guidelines, and assisting with annotation tools such as CVAT and Label Studio. Engage in basic data augmentation to enrich training data.
– Monitor and record experimental results using key metrics like mean Average Precision (mAP), recall, frames per second (FPS), and GPU utilization. Thoroughly document insights and progress to inform ongoing development.
– Collaborate closely with the team to debug and resolve issues encountered in field deployments. Participate in prototyping efforts and implement quick fixes as needed.
– Adhere to best practices in secure handling and processing of sensitive enterprise video data, ensuring compliance with organizational and industry standards.
**Qualifications and Skills**
– Proficient programming skills in Python, with hands-on experience in libraries such as NumPy, OpenCV, and frameworks including PyTorch or TensorFlow.
– Strong foundational understanding of computer vision and machine learning concepts, including convolutional operations, Intersection over Union (IoU), Non-Maximum Suppression (NMS), data augmentation techniques, overfitting, and evaluation metrics.
– Prior exposure to at least one specialization within computer vision, such as object detection, image segmentation, or object tracking.
– Demonstrated ability to independently learn and implement concepts by studying technical documentation, research literature, and developing supporting tools or utilities.
**Preferred (Bonus) Skills and Experience**
– Familiarity with the NVIDIA software ecosystem, including CUDA programming, TensorRT optimization, and DeepStream SDK.
– Practical experience working with NVIDIA Jetson platforms (Xavier, Orin) or other edge computing devices.
– Knowledge of multimedia frameworks like GStreamer or FFmpeg, including video codec standards.
– Experience with machine learning lifecycle tools such as DVC (Data Version Control), Weights & Biases (W&B), Docker containerization, or automated ML pipelines.
– Interest or experience in advanced computer vision domains such as re-identification (re-ID), human pose estimation, optical character recognition (OCR), or multi-camera video systems.
– Previous involvement in projects related to retail environments, manufacturing floors, warehouses, or surveillance applications will be considered advantageous.
**Educational Background**
– Candidates should be pursuing or have completed a B.Tech, B.E., M.Tech, or equivalent degree in Computer Science, Electrical Engineering, Mathematics, or related disciplines. Equivalent practical experience in relevant fields will also be considered.
**Why Join Us?**
At Agrex.ai, you will be part of a fast-paced, innovative environment where your contributions directly impact our state-of-the-art video analytics platform. This internship is an excellent platform for learning, professional growth, and gaining valuable exposure to deploying AI solutions in enterprise scenarios. If you are passionate about computer vision and are eager to engage with complex, real-world problems, we encourage you to apply and grow with our talented team.
