ML Engineer (Research Scientist - Cryosphere)

Arctic Research and Development; Cirencester; Hybrid; GBP 90,000 - 120,000; Full time

Benefits offered by this job

Competitive compensation

Private Health Insurance

Dental & Optician

Equity

Real field exposure

Mission-driven culture

Small team, real ownership

Job summary

Arctic Research and Development (ARD) is building Auka — an intelligence layer for the Arctic that turns observations into understanding, foresight, and action. You will own the Analyse-and-ML workstream, setting the direction for physics and ML collocation within ARD and helping found the science team.

This first science hire will collaborate with the Science Lead and Head of Software, shaping end-to-end pipelines and ensuring models ship to production while balancing scientific rigor and

Qualifications

• PhD in physical, geophysical, computational, or ML discipline with physics–ML interface work.
• Fluency with uncertainty quantification and validating models against real measurements.
• GIS or geospatial data handling
• Machine learning or data science background
• Strong scientific software engineering: production-quality Python, reproducibility, version control, testing
• Seniority to own a workstream and set direction independently

Responsibilities

• Distil reference physics into fast operators - emulators / surrogates that run orders of magnitude faster than reference physics models.
• Own the forward-to-inverse bridge - build ML that inverts hyperspectral and multispectral imagery into physical quantities with physics as constraint.
• Make trust a property of the model - bake in uncertainty quantification, provenance, and validation against cold-lab and field measurements.
• Set the technical bar for physics ⇄ ML across the science team, and mentor hires.
• Ship into the platform - land models in production with Science Lead and Head of Software.

Skills

Physics–ML hybrid

Uncertainty quantification

Geospatial data handling

Machine learning

Scientific software engineering

Independent leadership

Education

PhD

Tools

Python

Xarray

Dask

Zarr

Git

Job description

LOCATION: CALMSDEN, CIRENCESTER, UK

FULL-TIME, HYBRID: 2 DAYS (MIN) ON-SITE, 3 DAYS REMOTE

THE LAST FRONTIER

ARD was founded on a paradox. The Arctic is both the next great economic frontier and one of the most hostile environments on Earth. Rich in resources and strategically vital, yet bone-chillingly cold, dark for months, and fundamentally inhospitable to humans. This paradox—an explosion of activity in a place that resists human presence—is why we exist.

Our solution is autonomy.

We build systems that operate reliably and intelligently in the Arctic, so that humans don't always have to.

Founded by record-holding pioneers with decades of polar experience, we take calculated risks on hard problems that matter. If you want to build and deploy ground-up technology with real-world impact, at an inflection point in history that won’t reappear, we’d like to talk.

THE ROLE

ARD is building Auka - an intelligence layer for the Arctic that turns a flood of observations into understanding, foresight, and action. Our approach is deliberately grounded in modern science. We develop and run physics models, collect empirical data in the field and laboratory and then use that knowledge to extract meaning from our diverse datasets. We avoid black boxes — the products we ship trace back to the physics and data that produced it.

This role focuses on developing and tuning machine learning pipelines that are fast, trustworthy, and traceable, ingesting diverse datasets with an emphasis on remotely sensed imagery. These pipelines will form the foundational products on which our longer-term Arctic intelligence platform will build.

This is the first science hire and a senior one. You will own the Analyse-and-ML workstream, working directly with the Science Lead, set the technical direction for how physics and ML meet at ARD, and become a pivotal founding member of the ARD science team.

Core Responsibilities

• Distil reference physics into fast operators - build emulators / surrogates that run orders of magnitude faster than our reference physics models.
• Own the forward-to-inverse bridge - Build the ML that inverts real hyperspectral and multispectral imagery into physical quantities (albedo, melt state, impurity loading), with the physics as the constraint, not an afterthought.
• Make trust a property of the model, not a report - bake in uncertainty quantification, provenance, and validation against cold-lab and field measurements. Outputs must be defensible pixel-to-decision.
• Set the technical bar for physics ⇄ ML across the science team, and help shape and mentor the hires that follow you.
• Ship into the platform - work with the Science Lead and Head of Software so your models land in production — not just in papers.

Requirements

We are looking for a scientist who lives on the seam between physical modelling and machine learning — someone whose instinct is to respect and accelerate physics, not replace it with an opaque network.

Essential

• A PhD (or equivalent research track record) in a physical, geophysical, computational, or ML discipline, with a body of work at the physics–ML interface: physics-informed / hybrid models, emulators or surrogate models, neural operators, differentiable simulation, or data assimilation with learned components.
• Fluency with uncertainty quantification and with validating models against real measurements.
• GIS or geospatial data handling
• Machine learning or data science background
• Strong scientific software engineering: production-quality Python, reproducibility, version control, testing. Your models are meant to ship.
• Seniority to own a workstream and be delegated to — able to set direction, make sound calls under ambiguity independently, and represent the science externally. Comfortable being the first, and helping build the team.

Strongly Preferred

• Working knowledge of the cryosphere (glaciology, ice-surface / snow, or sea-ice physics). We will weight the ability to reason fluently about physical models over any specific sub-domain, and we also highly value marine and atmospheric science — but you should be able to engage with radiative-transfer and ice-surface physics without hand-holding.
• Experience with Earth-observation / remote-sensing retrieval — inverting satellite or drone imagery into physical quantities, atmospheric correction, cal/val.
• The cloud-native geospatial stack (Xarray, Zarr, Dask, STAC; the Pangeo ecosystem).

What We Value

• You reach for the simplest model that respects the physics before the biggest one.
• You treat provenance and uncertainty as first-class, not documentation added at the end.
• You've shipped something real that others depend on — an open-source tool, an operational model, a product.

Why ARD

The mission

- The Arctic is the fastest-changing region on Earth. You'll build the intelligence layer that helps understand and steward it — an early-warning system for the planet's coldest edge.

Science and product

- Our model rewards work that yields both new science and a platform others build on. You won't have to choose between publishing and shipping.

Foundational scope

- You're the first science hire, you'll work directly with the Science Lead and Head of Software. You'll be instrumental in shaping the team that follows.

Open by design

- A free Arctic data commons alongside paid services — your work reaches the whole field, not a walled garden.

What we offer

• Competitive compensation
• Private Health Insurance
• Dental & Optician
• Equity — meaningful options as an early employee at an early-stage company
• Real field exposure — travel to Arctic sites and Outposts when needed (genuinely, not as a gimmick)
• Mission-driven culture — focus on impact, not hours logged
• Small team, real ownership — what you build matters and ships