Applied Researcher, Audio Post-Training

About the role

Join our team to shape how users interact with our generative audio models. You will bridge the gap between research and real-world application, managing the entire lifecycle of model refinement. This role focuses on turning user feedback into technical improvements through data processing, fine-tuning, and evaluation.

Key facts

What you'll do

• Partner with product teams to translate user requirements into research objectives.
• Define research strategies to address ambiguous model behavior issues.
• Conduct experiments across the modeling stack, including synthetic data generation, supervised fine-tuning, reinforcement learning, and evaluation frameworks.
• Perform root cause analysis on production model failures to inform future iterations.
• Determine when specific model capabilities meet the standard for public release.

Requirements

• Proficiency in machine learning, software engineering, and debugging complex systems.
• Experience managing the quality and construction of large multilingual datasets.
• Practical background in training and troubleshooting generative models, specifically regarding SFT, RL, synthetic data, and automated or human evaluation.
• Ability to prioritize customer-driven problem solving over benchmark optimization.

Nice to have

• Native proficiency in languages other than English.

Skills & tools

• Supervised Fine-Tuning (SFT)
• Reinforcement Learning (RL)
• Synthetic Data Generation
• Model Evaluation (Human/Automated)
• Multilingual Dataset Management

Practical notes

• This is an on-site role.
• Visa sponsorship is available and evaluated on a case-by-case basis.
• Cartesia participates in E-Verify for all new hires.
• US benefits include fully covered medical, dental, and vision insurance, 401(k), commuter stipend, flexible PTO, daily meals, and parental leave (9 weeks paternity/12 weeks maternity).