Agrex Technologies Private Limited

About Agrex.ai Agrex.ai is a video analytics company transforming existing CCTV infrastructure into intelligent, real-time monitoring systems for retail, manufacturing, banking, logistics, and education. Our platform delivers people & vehicle analytics, SOP/compliance monitoring, and operational insights—on edge devices (e.g., Jetson) and in the cloud. Role Overview We’re looking for a Computer Vision Engineer (1–4 years) who’s excited to ship production-grade CV features on real video streams (RTSP/NVR/VMS). You’ll train and optimize models, stitch them into pipelines (DeepStream/GStreamer), and help deploy on Jetson Orin/Xavier and NVIDIA GPUs—working closely with Product, Solutions, and Support. What You’ll Do Implement, train, and evaluate detection/segmentation/tracking models (e.g., YOLO/RTDETR/Mask models, ByteTrack/OC-SORT). Integrate models into NVIDIA DeepStream / GStreamer pipelines; export to ONNX/TensorRT, optimize for FPS, latency, memory (FP16/INT8). Build features for people/vehicle counting, dwell/queue analytics, intrusion/zone rules, PPE/SOP checks, and basic re-ID. Work with streaming video: RTSP/H.264/H.265, jitter handling, reconnection logic, timestamp sync, and frame sampling. Write clean, testable Python (and some C++ for performance-critical parts); add unit/integration tests and simple benchmarks. Curate datasets: data selection, labeling guidelines, CVAT/Label Studio, augmentation, and dataset/version control (DVC/W&B). Run experiments, track metrics (precision/recall/mAP/F1, FPS, GPU util), and document results and tradeoffs. Debug field issues with Support/Solutions; contribute to root-cause analysis and hotfixes. Follow best practices for security & privacy with enterprise video data. Minimum Qualifications 1–4 years experience (strong final-year project/internships count). Solid coding in Python (NumPy, OpenCV, PyTorch/TensorFlow); basic Linux, Git, and debugging skills. Understanding of core CV/ML: convolutions, receptive fields, NMS/IoU, augmentation, overfitting, evaluation metrics. Familiarity with at least one: object detection, segmentation, or multi-object tracking. Comfort reading docs, reproducing baselines, and writing small utilities/tools. Nice to Have Exposure to NVIDIA stack: CUDA, TensorRT, DeepStream, Nsight profiling. Experience on Jetson (Xavier/Orin) or other edge devices. Knowledge of GStreamer, FFmpeg, video codecs/containers. Basics of MLOps (DVC/W&B/Mlflow), Docker, simple CI. Re-ID, pose estimation, OCR, or multi-camera stitching experience. Domain exposure to retail floors, factories, warehouses, or VMS integrations. Education B.Tech/BE/M.Tech in CS/EE/Maths or equivalent practical experience Job Type: Full-time Pay: ₹20,000.00 - ₹50,000.00 per month Benefits: Flexible schedule Leave encashment Paid time off Ability to commute/relocate: Dwarka, Delhi, Delhi: Reliably commute or planning to relocate before starting work (Required) Experience: PyTorch: 1 year (Required) NVIDIA Jetson: 1 year (Preferred) Deep Learning: 1 year (Required) Work Location: In person

About Agrex.ai Agrex.ai is a video analytics company transforming existing CCTV infrastructure into intelligent, real-time monitoring systems for retail, manufacturing, banking, logistics, and education. Our platform delivers people & vehicle analytics, SOP/compliance monitoring, and operational insights—on edge devices (e.g., Jetson) and in the cloud. Role Overview We’re looking for a Computer Vision Engineer (1–4 years) who’s excited to ship production-grade CV features on real video streams (RTSP/NVR/VMS). You’ll train and optimize models, stitch them into pipelines (DeepStream/GStreamer), and help deploy on Jetson Orin/Xavier and NVIDIA GPUs—working closely with Product, Solutions, and Support. What You’ll Do Implement, train, and evaluate detection/segmentation/tracking models (e.g., YOLO/RTDETR/Mask models, ByteTrack/OC-SORT). Integrate models into NVIDIA DeepStream / GStreamer pipelines; export to ONNX/TensorRT, optimize for FPS, latency, memory (FP16/INT8). Build features for people/vehicle counting, dwell/queue analytics, intrusion/zone rules, PPE/SOP checks, and basic re-ID. Work with streaming video: RTSP/H.264/H.265, jitter handling, reconnection logic, timestamp sync, and frame sampling. Write clean, testable Python (and some C++ for performance-critical parts); add unit/integration tests and simple benchmarks. Curate datasets: data selection, labeling guidelines, CVAT/Label Studio, augmentation, and dataset/version control (DVC/W&B). Run experiments, track metrics (precision/recall/mAP/F1, FPS, GPU util), and document results and tradeoffs. Debug field issues with Support/Solutions; contribute to root-cause analysis and hotfixes. Follow best practices for security & privacy with enterprise video data. Minimum Qualifications 1–4 years experience (strong final-year project/internships count). Solid coding in Python (NumPy, OpenCV, PyTorch/TensorFlow); basic Linux, Git, and debugging skills. Understanding of core CV/ML: convolutions, receptive fields, NMS/IoU, augmentation, overfitting, evaluation metrics. Familiarity with at least one: object detection, segmentation, or multi-object tracking. Comfort reading docs, reproducing baselines, and writing small utilities/tools. Nice to Have Exposure to NVIDIA stack: CUDA, TensorRT, DeepStream, Nsight profiling. Experience on Jetson (Xavier/Orin) or other edge devices. Knowledge of GStreamer, FFmpeg, video codecs/containers. Basics of MLOps (DVC/W&B/Mlflow), Docker, simple CI. Re-ID, pose estimation, OCR, or multi-camera stitching experience. Domain exposure to retail floors, factories, warehouses, or VMS integrations. Education B.Tech/BE/M.Tech in CS/EE/Maths or equivalent practical experience Job Type: Full-time Pay: ₹20,000.00 - ₹50,000.00 per month Benefits: Flexible schedule Leave encashment Paid time off Ability to commute/relocate: Dwarka, Delhi, Delhi: Reliably commute or planning to relocate before starting work (Required) Experience: PyTorch: 1 year (Required) NVIDIA Jetson: 1 year (Preferred) Deep Learning: 1 year (Required) Work Location: In person