Senior Machine Learning Engineer at hum.ai

Job description

SENIOR MACHINE LEARNING ENGINEER Location: San Francisco ABOUT HUM.AI hum.ai Hum.ai hum.ai is building planetary superintelligence. Backed by top funds, we’ve raised $10M+ and are now heads down building. Join us at the cutting edge, where we’re scaling generative transformer diffusion models, designing next-gen benchmarks, and engineering foundation models that go far beyond LLMs. You’ll be at the core of a moonshot journey to define what’s next in agentic AI and frontier model capabilities. We are looking for an experienced Senior Machine Learning Engineer who is eager to advance the frontier of AI, help us design, build, and scale end-to-end novel foundation models, and leverage their hands-on experience implementing a wide range of pre-training and post-training models, including large foundation models (beyond just LLM fine-tuning). This role is focused on: - Designing, implementing, and scaling state-of-the-art models - Productionizing research codes, models and technologically complex systems - Shaping benchmark design and model evaluation frameworks - Building agentic AI capabilities and long-term technical bets WHO ARE WE? Hum is a seed-funded startup on a mission to create positive impact through earth observation and AI. Founded at the University of Waterloo by a team of PhDs and engineers, we’re backed by some of the best AI and climate tech investors like HF0, Inovia Capital and Propeller Ventures, angels like James Tamplin (cofounder Firebase) and Sid Gorham (cofounder OpenTable, Granular), and partners like Amazon AWS and the United Nations. WHAT DO WE DO? We’re building multimodal foundation models for the natural world. We believe there’s more to the world than the internet + more to intelligence than memorizing the internet. Our models are trained on satellite remote sensing and real world ground truth data, and are used by our customers in nature conservation, carbon dioxide removal, and government to protect and positively impact our increasingly changing world. Our ultimate goal is to build AGI of the natural world. ABOUT THE ROLE The role will involve: - Collaborating with researchers and scientists to implement, evaluate and scale proof-of-concept models. - Owning, implementing and integrating the latest state-of-the-art methods and external open-source codes. - Develop AI systems capable of accurately understanding the universe and generating new knowledge. - Training multi-modal models supporting different sensor and other modalities like text REQUIREMENTS - Bachelor’s degree in computer science, engineering, a related field, or equivalent experience. - 5+ years of relevant work experience. - Prior experience building distributed training pipelines for multi-node systems using PyTorch and Ray. - Experience training large diffusion or transformer models. Preferably on video or time series data. - Proficiency with Python, Ray Trainer, PyTorch, and Anyscale framework. - Familiarity with cloud platforms such as AWS, GCP, or Azure. NICE TO HAVE - Past training of video or time-series models - Startup experience, comfortable with a small dynamic team. - Location wise, strong preference for in-person in Waterloo or San Francisco however remote work is possible for exceptional candidates.

Requirements

Must-haves

Location: San Francisco
Hum.ai http://Hum.ai is building planetary superintelligence. Backed by top funds, we’ve raised $10M+ and are now heads down building.
Designing, implementing, and scaling state-of-the-art models
Productionizing research codes, models and technologically complex systems
Shaping benchmark design and model evaluation frameworks
Building agentic AI capabilities and long-term technical bets
WHAT DO WE DO?
Collaborating with researchers and scientists to implement, evaluate and scale proof-of-concept models.
Owning, implementing and integrating the latest state-of-the-art methods and external open-source codes.
Develop AI systems capable of accurately understanding the universe and generating new knowledge.
Training multi-modal models supporting different sensor and other modalities like text
Bachelor’s degree in computer science, engineering, a related field, or equivalent experience.
5+ years of relevant work experience.
Prior experience building distributed training pipelines for multi-node systems using PyTorch and Ray.
Experience training large diffusion or transformer models. Preferably on video or time series data.
Proficiency with Python, Ray Trainer, PyTorch, and Anyscale framework.
Familiarity with cloud platforms such as AWS, GCP, or Azure.

Nice-to-haves

Past training of video or time-series models
Startup experience, comfortable with a small dynamic team.
Location wise, strong preference for in-person in Waterloo or San Francisco however remote work is possible for exceptional candidates.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

machine learningaigollmawspythongcpazure

Platform: ashby
Effort: quick