Earth Observation AI Research Scientist (Remote)

hum-ai; United States; Remote; USD 180,000 - 230,000; Full time

Benefits offered by this job

Equity

Health insurance

Dental & vision

Job summary

Hum is seeking an AI Research Scientist in San Francisco or remote to lead the center model in our Earth observation platform. You will collaborate with our Chief Scientist and engineers to train, evaluate, and operationalize models that answer customer questions and advance the world model.

You'll publish research on earth observation foundation models, participate in conferences, and help shape the product roadmap with milestones and risks, while scaling experiments from prototypes to

Qualifications

• PhD in computer science, engineering, or related field.
• 2+ years of industry experience building and shipping AI systems.
• Strong understanding of cutting-edge AI research with publications or projects.
• Experience scaling from small experiments to large-scale training.
• Comfort designing models for unconventional data modalities beyond text.

Responsibilities

• Own the model architecture and evaluate current approaches.
• Apply state-of-the-art techniques to diverse datasets with domain experts.
• Define success metrics beyond existing benchmarks.
• Run fast, cheap experiments and scale winners to large-scale training.
• Collaborate with the agent team to integrate the model into products.
• Publish research and present at conferences and to the community.

Skills

AI research

Publications

Experimentation

Education

PhD in CS/Engineering/related field

Tools

Python

PyTorch

Job description

Hum is seeking an AI Research Scientist in San Francisco or remote to lead the center model in our Earth observation platform. You will collaborate with our Chief Scientist and engineers to train, evaluate, and operationalize models that answer customer questions and advance the world model.

You'll publish research on earth observation foundation models, participate in conferences, and help shape the product roadmap with milestones and risks, while scaling experiments from prototypes to