The role, as published
About the company
Our client is an AI technology company.
The role
Raydar is recruiting for this role on behalf of our client. Own the experimentation cycle for improving machine learning models, from preparing data and defining evaluation criteria through training and handing results to production. Collaborate across engineering and leadership, and convert successful experiments into repeatable internal methods.
What you'll do
- Create and run training workflows that blend several learning techniques, and assess how each affects model quality and cost.
- Assemble training datasets generated from real usage, including choosing the source models and filtering the output.
- Build testing setups, scoring components and benchmarks so that gains can be measured credibly.
- Work with infrastructure engineers to reduce the compute and memory cost of experiments and keep training stable.
- Move successful research into live deployments and publish what you learn through articles, papers and open-source code.
What we're looking for
- At least 3 years of relevant experience.
- Practical background in model training or fine-tuning, gained in industry or an academic lab, with deep expertise in at least one technique.
- Understanding of how PyTorch or JAX run training jobs, and comfort debugging unfamiliar research code and confirming it behaves as designed.
- Sound scientific rigor: sensible baselines, fair comparisons and awareness of overfitting, contamination and flawed metrics.
- Clear, specific communication about your own contributions to past work and the evidence supporting your conclusions.
Benefits
- Health insurance and standard benefits
Location and work model
- San Francisco, CA, United States
- On-site
- Full-time