Computer Vision Engineer — Marine Habitat Monitoring

David Mensah

Full-time · Mid-level · London

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

We build underwater camera and sensor systems for marine ecology research and offshore infrastructure inspection. Our customers include marine research institutes, environmental monitoring agencies, and offshore energy operators who need to survey seabed habitats, pipeline integrity, and aquaculture infrastructure. The computer vision problem in underwater environments is distinctly harder than terrestrial vision: variable turbidity, colour distortion from water absorption, particle scatter, bio-fouling on camera housings, and the complete absence of labelled training data for the specific species and habitats our customers are monitoring. We are not a large company — 28 people — but the problem is technically interesting and the application domain is meaningful. We are hiring a mid-level computer vision engineer to join our software team as we expand our automated analysis capabilities. The role is fully remote but you will need to be in a European time zone — most of our research partners and clients are in Europe and we run collaborative sessions during European business hours.

Responsibilities

Develop object detection and segmentation models for species identification and habitat mapping from underwater footage
Build data augmentation and transfer learning strategies that work effectively when the annotated training set is measured in hundreds rather than millions of images
Contribute to the image pre-processing pipeline that corrects for underwater colour distortion and scattering artefacts
Collaborate with marine ecologists to understand classification schemes and evaluation criteria that are ecologically meaningful
Document models and evaluation methodology for scientific partners who need to understand and validate the analysis in their own research context

Requirements

4+ years of computer vision engineering with experience on challenging real-world image datasets — medical imaging, satellite imagery, industrial inspection, or underwater data are all relevant
PyTorch for developing and training object detection and segmentation models on small, domain-specific datasets where transfer learning and data augmentation strategies matter more than raw compute
Python at a professional engineering standard — reproducible training pipelines, clean evaluation code, and the discipline to document what you built for a small team that cannot afford undocumented magic
Scikit-learn for baseline model development and the evaluation framework that everything else is measured against
Docker for reproducible environments deployable to our edge computing hardware on underwater platforms
MLflow for experiment tracking and model versioning — with a small dataset, complete reproducibility of every result is non-negotiable
Curiosity about marine ecology and underwater imaging is genuine — the people who do best in this role are the ones who find the domain interesting, not just technically tractable

Benefits

€62,000 – €78,000 annual salary depending on experience
Fully remote — European time zone required
28 days annual leave
Annual budget of €1,500 for conferences, courses, or training
A technically unusual problem in a domain where the data doesn't look like anything in ImageNet — if that appeals to you, it is a genuine advantage