Applied Research Scientist, AI Research

Location: San Francisco, United States; hybrid or remote

Base salary: $261,625–$299,000 per year, plus equity and benefits.

Company

Descript is building an AI-powered video and audio editing tool for creators.

What you will do

• Build generative video and audio models for Video Regenerate, lip-sync, video translation, and other production features.
• Develop zero-shot voice and roomtone cloning from short reference recordings.
• Build multimodal vision-language systems and evaluation methods for agentic editing features.
• Work on computer-vision problems including digital human reconstruction and facial modeling.
• Research algorithms for media synthesis, speech enhancement, anomaly detection, and audio/video tagging.
• Identify research directions and take promising ideas from experiments to shipped product features.

Requirements

• Design and implement deep learning algorithms demonstrated through publications, open-source work, or shipped models.
• Strong programming skills and deep fluency in PyTorch and/or TensorFlow.
• Ability to generate, test, and evaluate machine learning ideas quickly with strong experimental judgment.
• PhD, Master's degree in deep learning or a related field, or equivalent experience.
• At least one of: lead or first authorship of an accepted publication at a top venue, or a key role in shipping a production feature centered on deep learning.
• Senior and Staff candidates should have experience owning research direction and mentoring or technically leading researchers or engineers.

Nice to have

• Experience with generative modeling for video, audio, or images.
• Experience with vision-language models and multimodal understanding.
• Experience in speech and audio modeling, including compression, enhancement, or synthesis.
• Experience building evaluations for generative or agentic outputs.
• Experience taking research ideas through to production-facing features.

Culture & Benefits

• Small, senior-heavy team with broad ownership from problem framing through experiments and analysis.
• Lean working style focused on clear technical communication and fast experimentation.
• Generous healthcare package and 401k matching program.
• Catered lunches and flexible vacation time.
• Remote work includes occasional opportunities for in-person collaboration; hybrid work offers flexibility around office attendance.