Research Engineer (Reinforcement Learning)
$135k - $300k • Remote • North America • FullTime
Posted 2mo ago
About the job
LiveKit is seeking an exceptional engineer to build post-training capabilities for their AI agents. These agents operate across voice and text channels, facing challenges related to long-term usefulness, context accumulation, and reliable tool usage during live conversations. The role involves developing training environments, owning the synthetic data pipeline, running end-to-end training experiments, and building evaluation frameworks to ensure model quality and performance.
Responsibilities
- Build environments and verifiers for model training.
- Own the synthetic data pipeline, including generation and quality control.
- Conduct end-to-end training experiments and analyze model performance.
- Develop evaluation criteria for model releases.
- Select and adapt open-weight base models for specific tasks.
- Ensure trained behaviors are robust for both voice and text agents.
- Deploy trained models into production and continuously improve them based on real-world usage.
Requirements
- Strong Python engineering skills.
- Experience taking a model from raw data to production.
- Treat data as a product, focusing on coverage, diversity, and leakage.
- Ability to design against weak rewards and anticipate model exploitation.
- Comfort with GPUs and an understanding of their limitations.
- Judgement on when to train models and when not to.
- Comfortable working collaboratively in a remote environment.
Benefits
- Competitive salary and equity package.
- Health, dental, and vision benefits.
- Flexible vacation policy.