Software Engineer, Python

Nearmap•Carlsbad, CA

1d•Hybrid

About The Position

As a Machine Learning Engineer, you will support a team of Data Scientists with tooling, data transformation, and deploying real-time models on Ray Serve endpoints. The team builds risk models and other related scores, using geospatial data as inputs to identify damage, rate the quality of a roof, or predict the likelihood of damage in the event of a natural disaster. Your role is to support the Data Science team's end-to-end workflow, and to be the key bridge between them and our ML Ops teams and systems in Australia and Poland.

Requirements

Around two years of industry experience writing Python in a professional software or ML engineering context
A degree in Computer Science or a related technical field
Hands-on experience working in a shared codebase: feature branches, pull request reviews, collaborative development with other engineers
Strong Python fundamentals and a track record of writing clean, maintainable, well-tested code
Solid data engineering skills, SQL, and experience with workflow tools like Airflow or Spark
Practical experience with LLMs, whether integrating APIs, fine-tuning, prompt engineering, or building LLM-powered components
Good communication, including the ability to talk to non-engineers without losing the technical plot

Nice To Haves

AWS experience (S3, EC2, ECS)
Docker and containerised environments
REST API integration at scale
MLOps and CI/CD
Familiarity with geospatial data

Responsibilities

Supporting the deployment of new risk models and scores in production.
Helping work with large scale data sets (hundreds of millions of rows), and building custom workflows on top of existing foundations.
Set up Claude Code usage patterns to allow team members to work more easily with AWS / Ray / Docker / Linux tooling.
Own data transformation pipelines that turn semantic geospatial maps of a property into attributes suitable for categorical modelling.
Collaborating with engineers to ensure solutions will work reliably on our tens-of-petabyte scale data sets, with multi-date, multi-angle, multi-modal data as inputs to algorithms.