Senior Computer Vision & Face Recognition Engineer

NearTech Search; United Kingdom; Hybrid; GBP 90,000 - 130,000; Full time

Job summary

NearTech Search is seeking a Senior Machine Learning Engineer to own a production-grade facial recognition and Computer Vision platform, from model development through deployment on NVIDIA edge hardware.

The role focuses on real-world deployment across detection, alignment, embeddings, and matching, balancing accuracy with performance. You will lead design and evaluation, collaborating with legal and data protection teams.

Qualifications

• 5+ years of production Computer Vision / ML experience.
• Experience deploying facial recognition systems using learned embeddings.
• Proficiency with PyTorch and ONNX/TensorRT for production deployment.
• Hands-on NVIDIA GPU, CUDA and inference optimisation.
• Experience with Docker, Linux, Git and CI/CD.

Responsibilities

• Build the full face recognition pipeline: detection, landmarks, alignment, embeddings, and matching.
• Train, fine-tune and evaluate embedding models.
• Develop evaluation frameworks for 1:1 verification and 1:N identification.
• Deploy and optimise models on NVIDIA edge GPUs.
• Work with NVIDIA DeepStream to maximise real-time video processing.
• Review and mentor mid-level engineers.

Skills

Computer Vision

Machine Learning

Facial Recognition

Python

C++

PyTorch

ONNX/TensorRT

NVIDIA GPUs

Docker

Linux

Git

CI/CD

Tools

DeepStream

CUDA

EDGE deployment

Job description

Senior Machine Learning Engineer - Computer Vision & Face Recognition

Location: UK Remote / Hybrid

Level: Senior - 5+ years' production Computer Vision / ML experience

The Opportunity

An exciting opportunity to take ownership of a production-grade facial recognition and Computer Vision platform, from model development through to deployment on NVIDIA edge hardware.

This is not a research-only role. You'll take models from development through to real-world deployment, working across face detection, alignment, embedding models, matching and evaluation, while balancing accuracy, performance and real-world constraints.

You'll have genuine ownership of the face recognition capability and the opportunity to shape how the technology is built, evaluated and deployed.

What You'll Do

• Build and develop the full face recognition pipeline - detection, landmarks, alignment, quality filtering, embeddings and matching.
• Train, fine-tune and evaluate face embedding models, including decisions around training data and loss functions.
• Develop evaluation frameworks for 1:1 verification and 1:N identification, including FAR/FMR, FNMR, TAR and threshold selection.
• Build multi-camera Computer Vision pipelines for real-world deployment.
• Deploy and optimise models on NVIDIA edge GPUs, using ONNX, TensorRT and FP16/INT8 quantisation.
• Work with NVIDIA DeepStream to maximise real-time video processing performance.
• Design gallery and enrolment systems, including indexing, similarity search, template management and thresholding.
• Work closely with data protection and legal teams to build responsible biometric systems, including retention, auditability and privacy requirements.
• Benchmark and track model performance across releases.
• Review technical work and mentor mid-level engineers.

What We're Looking For

• 5+ years' production Computer Vision / Machine Learning experience.
• Proven experience building and deploying facial recognition systems using learned embeddings.
• Strong understanding of face detection, landmark alignment and embedding approaches such as ArcFace, CosFace or AdaFace.
• Strong understanding of 1:1 verification vs 1:N identification and biometric evaluation metrics.
• Strong Python, with C++ experience or willingness to work with it.
• PyTorch for training and ONNX/TensorRT for production deployment.
• Hands-on NVIDIA GPU, CUDA and inference optimisation experience.
• Docker, Linux, Git and CI/CD experience.
• Comfortable making technical decisions based on measurable accuracy and performance.

Particularly Relevant Experience

• Facial recognition using CCTV or challenging real-world imagery, including low-resolution, off-angle faces, motion blur, occlusion and difficult lighting.
• NVIDIA DeepStream, GStreamer or Triton.
• Face quality assessment and template fusion.