Machine Learning Engineer

• Architect, train, and iterate on deep learning models purpose-built for complex scientific domains
• Push model performance forward through GPU optimization and distributed training strategies
• Ingest, validate, and transform massive geospatial datasets into clean, analysis-ready form
• Build and own scalable data pipelines that move high-volume scientific data reliably end to end
• Partner closely with domain scientists to embed ML into advanced simulation workflows
• Write production-grade code: tested, readable, and built to last
• Debug hard problems across distributed systems when they arise, and figure out why they happened

• 6+ years of software engineering experience, with meaningful time spent on ML systems
• Python fluency is required; proficiency in C++, Java, or Rust is a strong plus
• Real experience with PyTorch and a solid grasp of modern neural network architectures
• Familiarity with scientific computing libraries (NumPy, SciPy) and time series methods
• Hands-on work with geospatial data processing and big data pipelines
• Cloud platform experience on AWS or Google Cloud Platform
• Working knowledge of distributed computing, tensor operations, and GPU performance tuning
• Full-stack exposure and comfort with databases are a plus

Who Thrives Here

This is an R&D-heavy environment with a lot of open questions and not a lot of playbooks. The engineers who do well here are self-directed, intellectually curious, and comfortable building in ambiguity. If you need a well-defined ticket queue to feel productive, this probably isn’’’’’’’’’’’’’’'’t the right fit. If you like hard problems and want room to actually own your work, keep reading.

Our client is a mission-driven AI startup building next-generation forecasting infrastructure at the intersection of deep learning and geoscience. Their models run across a heterogeneous compute stack, public cloud, dedicated GPU clusters, edge-deployed hardware, and national supercomputing facilities, and their work directly supports federal defense and public safety missions. If you want your code to matter in the real world, this is worth a look., Our client runs a small team, roughly five full-time engineers plus contractors, led by a hands-on CTO. Everyone works across the full customer portfolio. They move fast but take code quality seriously, and they believe the best time to iterate is when the system is working, not when it’’’’’’’’’’’’’’'’s on fire.