Job Description

This role is all about translating R&D from other parts of the Nearmap AI & Computer Vision team into data and ultimately, answers. You will be a core contributor to the algorithmic systems that create the Map Data behind Nearmap's AI products. Every time Nearmap flies a survey, we find and measure what's on the ground (roofs, trees, roads and much more) and turn it into foundational map data: clean, property-level data that other products build on. We work with 2D and 3D aerial imagery captured daily across whole countries, alongside property boundaries, licensed open data and other geospatial sources. We get that data into the hands of the people and systems that rely on it, and keeping it affordable and trustworthy only gets more interesting as the volume grows.

We design, build, and operate software systems that take petabytes of data to transform aerial imagery to insight. The team focuses on algorithmic systems, where each node of the computational graph is a software application in its own right (such as a deep learning model or geospatial algorithm), and the goal is to produce outstanding high-quality data for customers and internal stakeholders in a fully automated and cost-effective manner. Our technology stack is based on the python scientific libraries and traverses machine vision deep learning technology such as Pytorch, and GIS tools such as the GEOS, Shapely and GeoPandas/GeoPolars libraries. We work mostly in python for speed of development and occasionally drop down to compiled libraries when we need to care about performance.

On a typical day, you might spend a morning tracking down why a pipeline slows to a crawl in one region but flies through another, and the afternoon with a machine learning researcher working out what it would take to run their new model over a whole country. You could finish the day reviewing a colleague's PR that makes inference faster, checking that the numbers still come out the same. Agentic coding tools like Claude Code are a normal part of how we work, and they're well suited to the exploratory, iterative work this role involves.

• Design, build and run production pipelines that turn aerial imagery and other geospatial data into property-level map data.
• 
  
• Keep those pipelines reliable: monitor runs, diagnose failures and fix root causes.
• 
  
• Make pipelines faster and cheaper to run through profiling, better use of GPUs and distributed compute, and compiled code where it pays off.
• 
  
• Measure the quality of everything we produce and validate every model or pipeline change against current outputs before it ships.
• 
  
• Turn research models into production systems that run across whole countries.
• 
  
• Develop geospatial algorithms, such as stitching features across tile boundaries and clipping them to property parcels.
• 
  
• Get data to the people and systems that use it, from batch deliveries to lower-latency services.
• 
  
• Review code and share knowledge so our scientific Python stays clean, tested and documented.
• 
  
• Build the tools and systems that let internal users find and use our data themselves and support them along the way.
• 
  
• 
  
• 
  

Qualifications

About You

• You're pragmatic. You value theory and best practice, but you'd rather ship something that works than build something elaborate.
• You're a team player. You communicate well, share what you know, and take good ideas from anyone in the team.
• You're eager to learn. You bring real-world experience and still expect to pick up new things.

What You'll Bring

Must Have

• Education: Bachelor’s in Computer Science, Engineering, Statistics, Physics or a related field (Master’s or PhD a bonus, not required)
• Experience: 5+ years in a numerical field such as Applied Mathematics, Physical Sciences, Computer Vision or Geospatial Analysis
• Production Code: proven ability to write high-quality production numerical code with a focus on business outcomes
• Tech Environment: Scientific Python, Linux, Git
• Software Engineering: modern best practice, clean code, robustness and correctness, building scalable ML systems
• Large Data: data too large for memory, computed across multiple nodes on scalable cloud compute
• Machine Learning: fundamentals including Regularisation, Hyperparameter Optimisation and Validation Methods
• Scientific Approach: forms hypotheses and validates them with statistical tests
• Mindset: strong Systems Thinking, balanced with pragmatism