Lead Computer Vision & Edge AI

Hawkvision AI; Jaipur; On-site; INR 2,400,000 - 4,800,000; Full time

Job summary

HawkVision AI in Jaipur, India seeks a hands-on Lead Computer Vision & Edge AI Engineer to own end-to-end model lifecycle, from data understanding to deployment across multi-camera CCTV streams. You will mentor engineers, establish scalable development, testing and deployment practices, and optimize inference pipelines using ONNX, TensorRT, DeepStream and GStreamer on NVIDIA RTX/Jetson.

This role focuses on production-ready solutions for safety analytics, with strong Python, PyTorch, and

Qualifications

• 510 years in machine learning or computer vision.
• 3+ years deploying CV solutions in production.
• Strong Python and PyTorch; YOLO or comparable detection frameworks; object detection, segmentation and multi-object tracking.
• Hands-on experience with real CCTV/RTSP video streams and NVIDIA GPU edge inference.
• Proficient with Linux, Docker, and edge deployment tools.
• Experience leading or mentoring engineers.

Responsibilities

• Design and improve object detection, segmentation, and tracking models for industrial CCTV environments.
• Data and training: label guidelines and dataset quality standards for rare safety events.
• Drive optimization for latency, throughput and edge computing constraints.
• Mentor computer vision engineers and establish scalable development practices.
• Translate customer safety requirements into scalable edge-AI solutions with product teams.

Skills

Python

PyTorch

Object Detection

Edge AI

Team Leadership

Tools

NVIDIA DeepStream

TensorRT

ONNX

GStreamer

Docker

Linux

CUDA

Job description

Location: Jaipur, Rajasthan

Work arrangement: Full-time, office-based

Experience: 510 years

Reporting to: CTO / CEO

About HawkVision AI

HawkVision AI is a UK-headquartered computer vision company with its engineering and delivery centre in Jaipur, India. We build real-time video analytics that help industrial organisations improve workplace safety and operational performance using their existing CCTV infrastructure.

Our platform covers use cases including PPE detection, vehiclepedestrian proximity monitoring, restricted-zone intrusion, person-down and fall detection, unsafe work-at-height, fire, smoke, spill and leak detection, and material-handling and behavioural safety analytics.

We have customers and pilots across India, the USA, the UK, Europe, the Middle East and Africa, and have been recognised by Verdantix as an innovative provider of video analytics for workplace safety. We are on our path to becoming the leading company in this space, working with Fortune 500 companies across the globe. HawkVision AI is ISO 27001 certified and SOC 1 compliant, and is currently working towards SOC 2 compliance.

The role

We are looking for a hands-on Lead Computer Vision & Edge AI Engineer to own the complete lifecycle of our computer vision models from understanding the safety problem and preparing datasets through training, validation and deployment of optimised inference pipelines across live multi-camera CCTV streams.

This is not a pure research or model-training role. You will also mentor our computer vision engineers and establish scalable development, testing and deployment practices.

Key responsibilities

• Model development Design, train and improve object-detection, segmentation, tracking and activity-recognition models for challenging industrial CCTV environments poor lighting, occlusion, unusual camera angles, low resolution and long distances.
• Solve small-object detection for PPE such as helmets, gloves and safety shoes, and build temporal models for person-down, falls and unsafe interactions.
• Select model architectures that balance accuracy, speed and edge-computing constraints.
• Data and training Own data-collection, annotation and dataset-quality standards, including labelling guidelines and balanced datasets for rare safety events.
• Drive iterative improvement through hard-negative mining and active learning; identify dataset bias, leakage and overfitting.
• Establish camera-wise and use-case-wise performance evaluation.
• Edge AI and production deployment Convert and optimise models with ONNX and TensorRT; build DeepStream and GStreamer pipelines on NVIDIA RTX and Jetson edge systems.
• Apply FP16/INT8 quantisation; optimise GPU memory, batching, decoding and multi-camera throughput; benchmark streams supported per GPU configuration.
• Diagnose frame drops, latency and RTSP instability in production; work fluently with Docker and Linux.
• Accuracy and performance management Set precision, recall and false-alarm targets, and measure real incident-detection performance rather than headline model accuracy.
• Analyse failures, build validation datasets that reflect real customer environments, define acceptance criteria for pilots and production, and automate regression testing.
• Technical leadership Mentor computer vision engineers and annotation teams; review architectures, training code and deployment pipelines; set coding, documentation and experimentation standards.
• Translate customer safety requirements into feasible solutions alongside product, delivery and customer teams; support technical discussions with enterprise customers; shape the long-term vision and edge-AI architecture of the platform.

Essential experience

• 510 years in machine learning or computer vision, including 3+ years deploying computer vision solutions in production.
• Strong Python and PyTorch; YOLO or comparable detection frameworks; object detection, segmentation and multi-object tracking.
• Hands-on experience with real CCTV/RTSP video streams and NVIDIA GPU edge inference.
• NVIDIA DeepStream, TensorRT, ONNX and GStreamer; good working knowledge of Linux and Docker.
• Proven optimisation of models for latency, throughput and GPU utilisation.
• Strong grounding in precision, recall, F1, confusion matrices and false-alarm analysis.
• Skilled at diagnosing data-quality, model-performance and production issues; writes production-quality, maintainable code.
• Experience leading or mentoring engineers.

Preferred experience

• Domain experience in industrial safety, manufacturing, logistics, construction, mining or CCTV analytics.
• PPE detection, behavioural safety analytics, person-down/fall or action-recognition models; small-object detection in heavily occluded scenes.
• CUDA or C++ optimisation; deployments across 20+ simultaneous video streams; NVIDIA Triton Inference Server.
• Model quantisation, pruning and .