Cartesia
Researcher, Post Training
$180K–$350K
*HQ - San Francisco, CA · On-site · Full-time · Posted Oct 21, 2025
- Visa sponsorship
- Visa sponsorship support is provided and assessed case by case; sponsorship is not guaranteed and depends on the role and work location.
About the job
Research and build post-training methods, evaluations, and systems that improve alignment, reasoning, and capabilities in multimodal models.
What you'll do
- Lead research initiatives to improve multimodal model alignment and capabilities, including developing post-training methods and evaluation frameworks.
- Build, debug, and scale experimental systems for reliable, reproducible training and evaluation.
- Collaborate across research, product, and platform teams to translate findings into production-ready specialized models.
What they ask for
- Deep knowledge of preference optimization and alignment methods, including RLHF and related approaches.
- Experience designing evaluations and metrics for generative or multimodal models.
- Strong engineering and debugging skills, including experience building or scaling complex ML systems.
- Ability to diagnose model-performance behavior across training and evaluation pipelines.
Nice to have
- Experience with multimodal model training, such as text, audio, or vision-language models.
- Contributions to alignment research or open-source evaluation or fine-tuning projects; experience designing or implementing human-in-the-loop evaluation systems.
Pay
$180K – $350K (base salary)
Tools
RLHF
AI focus
Develop post-training and alignment methods for multimodal models, using preference optimization, evaluation, and feedback-driven learning to improve reasoning, consistency, and alignment with human intent.
Model-written summary · the original posting is authoritative