Machine Learning Engineer, Model Experimentation

Aether Biomachines•South San Francisco, CA

1d•$150,000 - $200,000

About The Position

Aether engineers enzymes to produce materials that are difficult or impossible to make by other means. Its print materials are already shipping into defense and unmanned systems, and the company is developing models that use protein-material binding data to identify proteins capable of making a desired material. The Machine Learning Engineer, Model Experimentation will rapidly test modeling ideas and provide evidence-based recommendations about which ideas merit further work. Working day to day with senior members of a small ML team, the engineer will run experimental cycles involving baselines, model changes, training, and evaluation. The work uses data from real lab measurements, allowing results to be checked against experimental outcomes. Success is measured by how quickly and clearly the engineer determines whether ideas hold up, rather than by the number of ideas tested. The role is suited to a hands-on experimenter who can implement ideas quickly, evaluate them honestly, communicate results clearly—including negative results—and stop experiments when evidence warrants it. No chemistry or biology background is required; Aether will teach the molecular and protein domain, as well as its data and tooling. Compensation is $150,000–$200,000 per year plus equity, with final compensation based on skills and experience. Aether welcomes applicants from all backgrounds, bases employment decisions on qualifications, merit, and business needs, and offers reasonable accommodations during recruitment.

Requirements

Hands-on experience training and evaluating models, gained through research, employment, competitions, or serious personal projects.
Strong experimental habits, including establishing baselines first, changing one thing at a time, performing ablations, and evaluating honestly.
Ability to work quickly and write code that runs promptly, with cleanup deferred until an idea proves worthwhile.
A track record of stopping experiments or projects, and ability to explain what evidence led to stopping and how quickly the decision was made.
Clear written communication of results, especially negative results.
Fluency in Python and a modern deep-learning framework such as PyTorch or JAX.

Nice To Haves

Experience with world models or latent-prediction methods, such as JEPA or DreamerV3.
Experience with diffusion models in any domain.
Competition results, open-source contributions, or a project built and maintained independently.

Responsibilities

Design and run machine-learning experiments end to end, including establishing baselines, making model changes, running training, and evaluating results.
Make a clear decision on each experiment and write up the result in a page or less, including failed experiments.
Build evaluations that can be trusted even when they contradict the idea being tested.
Bring ideas from the wider field, including world models, latent-prediction objectives, and diffusion, and test them against Aether's data.
Work within the team's ML infrastructure and improve it where it slows experiments down.
Work day to day with senior ML team members, who help set priorities and review experimental decisions.