About the company
Our client is a software company.
The role
Raydar is recruiting for this role on behalf of our client. Drive applied machine learning research on language models, moving from an initial question through experimentation to a model ready for real-world use. This is an early technical position with close collaboration alongside company leadership, engineers and clients.
What you'll do
- Convert client workflows and observed model shortcomings into focused research questions, evaluation suites and experiments.
- Create training datasets, reward functions and automated checkers, and run fine-tuning, preference-based and reinforcement learning experiments.
- Investigate failed training runs and rely on ablation studies and held-out testing to learn which changes truly raise quality.
- Lead projects from start to finish, ending with a deployed model or a publishable result.
- Collaborate with fellow researchers and infrastructure engineers to ship successful models and study how they behave live.
Requirements
What we're looking for
- Significant practical background in adapting language models after pretraining, with personal accountability for experiments and their outcomes.
- A track record where your modeling work reached real use or outside review, whether through a shipped product, a widely adopted open model, or an accepted paper.
- Published research or open-source machine learning contributions in this area.
- Prior experience at an organization operating at the leading edge of AI.
- Strong Python and PyTorch skills, including writing and debugging training code on your own.
- Good analytical judgment, with the ability to explain why a method succeeded and where a metric gave a false signal.
- Confidence working through ambiguous technical problems and speaking directly with clients and colleagues.
Bonus points
- Experience mentoring more junior engineers or researchers.
- Master's or PhD in Computer Science.
- Background in areas such as long-horizon reinforcement learning, model compression, generated training data, reward design or large-scale distributed training.
Benefits
Compensation and benefits
- Base salary: USD 200,000 to 350,000 per year
- Equity
Location and work model
- Mountain View, CA, United States
- On-site, 5 days a week in the office
- Full-time