Founding Geospatial ML Engineer

About Aureum Aureum is bringing satellite intelligence to... literally everyone. Right now, satellite intelligence is one of the most powerful information sources on the planet, and yet almost nobody can use it. It is priced, gated and skilled out of reach of the investors, analysts, journalists and organisations who need it most. We think that is absurd, and we are fixing it: taking a capability that today requires a specialist team and a six-figure budget, and putting it in anyone's hands. We are an early-stage startup with a small founding team, the beta launches in the next few weeks, and this is only just the beginning. What you'll do You will own the intelligence core of the platform: everything between the satellite and the answer. It already exists, but you'll be taking ownership of it off the CEO's plate. Getting the data in. You will own how imagery and complementary data flow into the platform from free and commercial satellite sources: building robust, scalable ingestion, making smart cost and quality trade-offs, and keeping the pipeline fast and reliable as we grow. Turning pixels into answers. You will own the analytical engine itself: applying computer vision and machine learning to extract real-world signal from satellite imagery, from detecting and counting objects to measuring change over time. You will choose the right technique for each problem, build it, evaluate it honestly, and improve it. Making the system smarter. You will own how the platform learns: how results are stored, reused and calibrated, and how the system gets better with every question it answers. This is a true founding role. You will have enormous ownership, direct influence on product and technical direction, and hard, genuinely novel problems to work on from day one. Everyone on the team is technical, and everyone lives for problem-solving. Come solve problems with us. Who you are You will need: Strong production Python and PyTorch, with real ML systems shipped, not just notebooks Applied computer vision depth: you have built and deployed models for detection, segmentation or change analysis against real-world imagery Experience fine-tuning pretrained models and evaluating them honestly The ability to build data pipelines end to end: APIs, storage, orchestration Evidence that you dive into unfamiliar domains and figure them out - more than anything else the ability to take an unfamiliar problem, think creatively about how to solve it, and then push relentlessly to make it happen Ideally, some of: The Earth observation stack: rasterio/GDAL, STAC, TorchGeo, Sentinel and Landsat data, map projections, radiometry Familiarity with geospatial foundation models and efficient fine-tuning methods MLOps: experiment tracking, model versioning, inference serving SAR, hyperspectral, or other non-optical modalities Deep geospatial or space-industry experience is genuinely valuable here, but it is not a requirement. What we care about most is raw problem-solving ability, intellectual honesty, and the drive to learn a new domain fast. Everyone at Aureum has taught themselves this industry from first principles; you will too. How we work Small team, high trust, high ownership. You own a whole domain, you argue what's most important to ship next, and then you ship it. Decisions are made on evidence, results are reported honestly (including the failures), and nobody is precious about how things have always been done. The CEO genuinely loves being told he's wrong about something (seriously). How to apply Send a CV or LinkedIn profile plus a short note on the most interesting problem you have solved with imagery or ML, and a link to anything you have built. GitHub repos, competition entries, papers and side projects all count equally. PhDs and top-tier schools are great, but what matters most is what you've done. Show us who you are and why you're right for this role - we'll listen.