JobCubby JobCubby
Sign up
S

Machine Learning Engineer

Security Level 5 San Francisco, California, United States · $200K–$350K/yr

Technology, Information and Internet · 2-10 employees

2 d ago
machine-learning Senior (5-10 yrs) Full-time United States
Log in to apply, save this posting, or score it against your profile with AI.

About the role

Design and conduct experiments to measure the impact of security controls on machine learning research and engineering workflows. Forecast future frontier workloads to ensure infrastructure and security controls are evaluated against realistic, large-scale AI requirements.

What they look for

Machine Learning AI Security Infrastructure Design Multi-node Jobs Accelerators Orchestration Storage Data Pipelines Experimental Design Frontier Models Training Stack Inference Fine-tuning Interpretability

Requirements

Candidates must have experience training models at an AI lab or equivalent and possess deep knowledge of the training stack including multi-node jobs and orchestration. You must be capable of designing experiments that provide actionable insights rather than just raw data.

Benefits

Healthcare Dental Vision 401k match

Full description

About Security Level 5

Security Level 5 is a Bay Area AI security nonprofit working with AI labs and US intelligence agencies to defend frontier models from priority nation-state-level threats. We move fast, and we're tripling headcount this year.

The Mission

We're hiring for what's plausibly the most difficult, urgent, and interesting challenge in AI security engineering this decade (for humans, at least): designing and building the first datacenter purpose-built to Security Level 5, the standard for defending frontier AI labs against attackers with the resources of a nation-state and billion-dollar budgets. We're working on the first SL5 components this year and aiming for an at-scale SL5 datacenter build in 2027.

We wrote the first version of the standard with input from CISOs and senior security staff at the frontier labs, and US security and intelligence officials. Now we're designing the SL5 architectures purpose-built for frontier AI workloads, and standing up the reference tech stack to guide labs' buildouts. This stack has to hold against nation-states, not destroy researchers' productivity, not cost more than labs can pay, and ship in 2-3 years.

What you'd work on

  • Designing experiments that measure the effect of SL5 controls on real ML research and engineering workflows.
  • Forecasting 2028 frontier workloads (training, inference, fine-tuning, interpretability) and the infrastructure they imply, so controls are evaluated against what labs will actually be doing.
  • Running the experiments at whatever scale is tractable and interpreting them for a frontier-scale audience.

You'd be a fit if you

  • Have trained models at an AI lab or equivalent, and are current on frontier methods and infrastructure.
  • Are comfortable across the training stack: multi-node jobs, accelerators, orchestration, storage, data pipelines.
  • Can design an experiment that answers a question rather than one that produces a number.
  • Are happy working at small scale as a proxy for frontier scale, and can reason about where the proxy breaks.

Logistics

Bay Area, in-person. This role's base salary range is $200,000 - $350,000 USD per year. We may be open to paying exceptional and/or highly experienced employees more than this. We also provide full benefits, including healthcare, dental, vision, 401k match, and more.

Similar roles