ESEN
← All open roles

Data Scientist (Machine Learning), mid-level

Detection and satellite · Remote · Full time

Go to the application →

You work on our own satellite detection models: improving them, measuring every change with data and taking to production what works.

Location

Remote, with no country restriction. We ask for at least 4 hours of daily overlap with Argentina time (GMT-3), because much of the work happens alongside the technical team. If you live in Buenos Aires, part of the work can be in person with the team.

The role

We are looking for a Data Scientist with a solid machine learning foundation to join a technical team that does a bit of everything: satellite detection, AI cameras, burned area and product. Today the main problem is the satellite detection pipeline, which runs in production across 21 countries on our own AI models. You will work on those models, measure every improvement with data, and take to production what works.

You report to the technical team lead and work alongside the detection team. The models running today were built by the team from scratch, so you will go deep into the architectures, not just consume them.

The central challenge is solving problems that have no established solution today. We are looking for an applied research mindset, where success is measured when the prototype ends up deployed and running in production.

The team is small and flat. Beyond satellite work, you will pick up other fronts when they are needed.

What you will do

  • Improve the satellite detection models: segmentation, super-resolution, classification and change detection. Depending on the case, tune what we already have, modify architectures or train from scratch.
  • Cut false positives and raise recall, with systematic measurement by region and by satellite source. Every model change is justified with data, not intuition.
  • Design and maintain reproducible benchmarks against ground truth (official perimeters, FEDS, FIRED, Copernicus EMS), with precision and recall reports by pipeline layer.
  • Improve perimeter and burned area calculation from satellite detections. It is the basis of the product we sell to insurers.
  • Build sellable data products for outside the fire season: deforestation monitoring, canopy degradation, NDVI and NDWI time series.
  • Produce detection reports for prospects using the team’s methodologies, and join client calls when the conversation gets technical.
  • Lend a hand on other fronts when the team needs it: AI cameras, pipeline infrastructure, integrations.
  • Document, keep the pipelines reproducible and do active code review with the rest of the team.

Requirements

  • A degree in computer science, systems or software engineering, or a related field. Advanced student or graduate.
  • Experience training, evaluating and tuning deep learning models on imagery with PyTorch or TensorFlow. Not only consuming pre-trained models: modifying architectures and training from scratch when needed.
  • Advanced Python, and code that can be maintained and deployed. Notebooks to explore, code to ship.
  • Experience designing benchmarks: defining metrics, ground truth and baselines, and reporting honest results even when they are not the ones you hoped for.
  • Having taken a model to production, or to something other people use, however small.
  • Technical English to read papers and write documentation.

Who we are looking for

  • You get on with it. Faced with a new problem, you build a prototype this week, measure it, and then decide whether it is worth perfecting.
  • You iterate fast and are not afraid to fail. An experiment that does not work and teaches you something is worth more than a month of planning.
  • You explain and defend your technical decisions with data. If someone asks why you chose A over B, you have an answer.
  • You are curious. You try different paths without being asked, and you bring proposals, not just deliverables.
  • You work in pairs. Pair programming, code review, asking for and giving feedback without fuss.
  • You use AI as part of how you work: coding agents, paper synthesis, exploratory analysis. It multiplies your output, it does not replace your judgment.
  • A small team where everyone touches everything suits you. There is no closed role waiting for you, and part of the work is defining it.
  • You care about the problem. You want to understand how fire behaves, why a satellite sees it or misses it, and what a firefighter does with the alert.

Bonus points

  • Verifiable publications, theses or open source work.
  • Experience working on research problems.
  • Experience with satellite imagery or geospatial data: rasterio, xarray, geopandas, GDAL, PostGIS.
  • Command of spectral indices (NDVI, NDWI, NBR) and raster and vector operations.
  • Having worked with sources such as MODIS, VIIRS, GOES, Sentinel-2/3, Landsat or Meteosat.
  • Fire data (perimeters, hotspots, FRP) or land use change data.
  • U-Net and variants, vision transformers or super-resolution on satellite imagery.
  • AWS (S3, Lambda, EC2) and serverless pipelines.

Our stack

Python (pandas, geopandas, rasterio, xarray, shapely), PostGIS, GDAL, Sentinel Hub, FIRMS API, LSA-SAF. ML with PyTorch, TensorFlow and scikit-learn; our own super-resolution and segmentation models trained from scratch on satellite imagery. AWS (Lambda, EC2, S3, API Gateway) with batch pipelines and updates every 10 minutes. GitHub and AI agents for development.

What we offer

  • Compensation in USD.
  • 15 working days of paid holiday.
  • A mission with real consequences. Your models produce detections that reach firefighters and reduce the damage in actual fires.
  • Satellite datasets at continental scale, and the freedom to experiment on them.
  • Objectives defined from day one, and a role and compensation review at 12 months against those objectives.
  • Remote work with flexible hours. A global team with hubs in the Americas and Europe.

What happens next

You fill in the form below. From there the process is six stages:

  1. Screening. We review your profile and your answers.
  2. Career interview. We walk through your career in detail, role by role.
  3. Challenge. A scoped exercise on the real work of the role.
  4. Interview with the CEO.
  5. References.
  6. Offer.