Location

Cambride, MA or remote

The challenge

Zeus AI is building foundation models for global and regional scale Earth system modeling, ingesting observations to produce an accurate low latency representation of the planet through machine learning data assimilation. This problem is often framed as observation to observation forecasting. We are tackling this problem with a multi-modal and multi-resolution modeling framework using numerous observation types including satellites, stations, aircraft, drones, and marine vessels to power forecasts and digital twin models. The machine learning architecture plays a key role in this system to learn relationships across sensors best characterizing the system’s state. 

The role

We're looking for a Machine Learning Scientist to join our core science and engineering team. In this role, you'll be instrumental in building foundation models for data assimilation and forecasting. You'll develop and fine-tune generative machine learning models for medium-range forecasting, working large-scale, heterogeneous Earth system datasets, rigorously evaluating model performance, and scaling training runs on large clusters. You'll apply these models to critical applications like weather prediction, aviation planning, air quality, renewable energy, and extreme weather.

Duties and responsibilities

• Advance the development of medium-range forecasting using EarthNet, Zeus AI’s foundation model for Earth observations
• Explore and adapt cutting-edge machine learning architectures, including Graph Neural Networks (GNNs), diffusion models, and autoregressive models for forecasting
• Carry out large-scale training runs in high performance computing environments
• Design and train data-specific compression algorithms to optimize the representation of heterogeneous Earth observation data for large foundation models
• Lead efforts to fine-tune models for downstream tasks like forecasting, air quality, atmospheric convection, and localized predictions, potentially utilizing super-resolution techniques or specialized architectures
• Collaborate closely with data engineers and domain scientists to integrate diverse Earth observation datasets into model training and validation pipelines.
• Engage in scientific communication including publications and conference presentations

Qualifications

• Ph.D degree in computer science, physics, mathematics, or related quantitative field
• 2+ years of industry experience as a software or machine learning engineer
• 2+ years of experience working with Earth science datasets including remote sensing datasets, weather forecasts, and/or climate simulations
• Excellent programming skills in Python including proficiency with deep learning frameworks like PyTorch or Tensorflow
• Experience with best practices in scientific software development, including version control and code optimization for high performance computing environments
• Proven experience implementing and training generative machine learning models including but not limited to variational autoencoders, transformers, and GNNs
• Strong interest in scientific research and development, with a desire to contribute to cutting-edge advancements
• Strong communication and collaboration skills to work effectively with an interdisciplinary team

What we offer

• The opportunity to work on cutting-edge research at the intersection of Earth science and machine learning, with access to extensive computing resources.
• A collaborative and stimulating work environment alongside a team of passionate and talented scientists and engineers.
• Competitive compensation, including salary, equity, and a comprehensive benefits package (health, dental, vision, and 401k retirement options).
• Flexible work arrangements and a co-working space benefit.
• The chance to make a tangible difference in the world by contributing to climate solutions and significantly improving weather forecasting.