Descript
Applied Research Scientist, AI Research
$197K–$262K
San Francisco, CA or Remote, US +2 · Remote · Full-time · Posted Aug 24, 2026
- Location
- San Francisco, CA or Remote, US
Remote
San Francisco, California, United States - Experience
- No specific number of years stated. At least one qualifying publication or production-shipping experience is required. Senior and Staff candidates have additional direction-setting and leadership expectations.
- Education
- A PhD or Master's in deep learning or a related field, or equivalent experience.
About the job
Develop specialized multimodal models that help Descript’s AI editing agent understand audio and video, evaluate editorial quality, and support production editing features.
What you'll do
- Build vision-language systems, datasets, and benchmarks for reasoning about visual and audio project content and editorial quality.
- Train or fine-tune specialized models and ship them to production with agent and engineering teams.
- Identify product-focused research directions; publishing is optional.
What they ask for
- Demonstrated ability to design and implement deep learning algorithms through publications, open-source work, or shipped models.
- Strong programming skills and deep fluency in PyTorch, with the ability to rapidly run and evaluate experiments.
- Strong experimental judgment and clear written and verbal communication.
- A PhD or Master's in deep learning or a related field, or equivalent experience; and either lead/first authorship at a top venue or a key role shipping a production feature using deep learning.
Nice to have
- Experience or depth in vision-language models, multimodal understanding, generative modeling, foundation-model post-training/fine-tuning/RL, or evaluation of generative or agentic outputs.
- For Senior and Staff candidates: experience owning research direction and mentoring or technically leading researchers or engineers.
Pay
$197,000–$262,500 USD base salary
Tools
PyTorch
AI focus
Build multimodal understanding capabilities for an AI editing agent, including vision-language systems, model training and fine-tuning, synthetic data, and evaluation of editorial quality.
Model-written summary · the original posting is authoritative