Principal AI Engineer – Computer Vision & Video AI

UMATR; Al Khobar; On-site; SAR 500,000 - 900,000; Full time

Job summary

UMATR is seeking a Principal AI Engineer to own the AI and perception stack for an AI-driven safety platform used in large-scale industrial environments. You will stay hands-on while setting architecture and solving advanced computer vision and multimodal AI challenges, influencing engineering direction.

The role centers on video-centric product development, including edge deployment, real-time systems and responsible AI with strong focus on performance, latency and safety controls.

Qualifications

• Proven experience taking multiple production ML or computer vision systems from concept through to live operation.
• Deep hands-on experience with video, streaming, analytics or computer vision over video.
• Experience with PyTorch or TensorFlow across detection, segmentation, tracking and video understanding.
• Production experience with vision-language or multimodal models, including evaluation and integration.
• Experience building agentic AI or tool-using systems with guardrails and human-in-the-loop controls.
• Strong dataset curation, fine-tuning, active learning and experiment management.

Responsibilities

• Own the technical direction of the computer vision and perception stack.
• Design and build production systems for object detection, segmentation, tracking and video understanding.
• Develop multimodal and vision-language systems that reason over live and recorded video.
• Build AI agents capable of reasoning over visual data using tools and APIs in controlled workflows.
• Own the full model lifecycle from problem definition to deployment, monitoring and improvement.
• Optimize models across edge and cloud, balancing accuracy, latency and cost.
• Establish responsible AI and safety controls including thresholds and auditability.
• Lead architecture reviews and document key technical decisions.

Skills

Production ML / Computer Vision

Video understanding

PyTorch / TensorFlow

Vision-language / Multimodal models

Agentic AI / Tool-using systems

Data curation / Active learning

Edge & cloud deployment

MLOps / Experiment tracking

English communication

Architecture design

Job description

We’re partnering with a technology company using AI to improve safety across large-scale industrial and construction environments. Their platform analyses live site video to identify potential hazards and provide teams with timely, actionable safety insights.

They’re looking for a Principal AI Engineer to take technical ownership of the AI and perception stack. You’ll stay hands‑on while setting architecture, solving the hardest computer vision and applied AI problems, and influencing technical direction across the wider engineering team.

This is a genuinely deep technical role where video is at the centre of the product, spanning computer vision, multimodal models, agentic AI, edge deployment and real‑time systems.

What You’ll Do:

• Own the technical direction of the computer vision and perception stack.
• Design and build production systems for object detection, segmentation, tracking and video understanding.
• Develop multimodal and vision-language systems that can reason over live and recorded video.
• Build AI agents capable of reasoning over visual data, using tools and APIs and operating within controlled workflows.
• Own the full model lifecycle from problem definition and data strategy through to experimentation, deployment, monitoring and continuous improvement.
• Optimise models across edge and cloud environments, balancing accuracy, latency, reliability and cost.
• Establish responsible AI and safety controls, including confidence thresholds, human review, explainability and auditability.
• Lead architecture reviews and document key technical decisions.

Who We’re Looking For:

• Proven experience taking multiple production ML or computer vision systems from concept through to live operation.
• Deep hands‑on experience working with video, including video streaming, analytics or computer vision over video.
• Strong experience with PyTorch or TensorFlow across detection, segmentation, tracking and video understanding.
• Experience with both purpose-built computer vision models and transformer-based approaches.
• Production experience with vision‑language or multimodal models, including model adaptation, prompting, evaluation and integration.
• Experience building agentic AI or tool‑using systems, with appropriate guardrails and human‑in‑the‑loop controls.
• Strong experience with dataset curation, fine‑tuning, active learning and experiment management.
• Proven experience deploying AI workloads across edge and cloud infrastructure, with a focus on real‑world latency, throughput and memory constraints.
• Strong MLOps and evaluation experience, including regression testing, experiment tracking, model versioning, monitoring and drift detection.
• Exceptional written and spoken English.

Nice to have:

• RAG, vector search or multimodal retrieval experience.
• Real‑time video or streaming camera systems.
• Edge AI technologies such as TensorRT, ONNX Runtime, Jetson or other GPU/NPU platforms.
• Self‑hosted LLM/VLM serving using tools such as vLLM or similar.
• Experience in safety‑critical, industrial or construction environments.
• Knowledge of EHS or workplace safety frameworks.
• Experience with privacy‑focused video systems, including anonymisation, access controls and audit trails.