Descript Tech Stack

AI-powered video editor for creators and teams built on text-editing paradigms

Software Development San Francisco, CA 51–200 employees Founded 2017 Privately Held

Descript is a video and podcast editing platform where text edits drive video cuts—a model that lowers the barrier to professional content creation. The tech stack reveals an ML-heavy architecture: PyTorch, TensorFlow, and CUDA for model inference, plus Spark and Dask for data processing at scale. Active projects center on third-party model integrations, audio synthesis, and MLOps infrastructure, while pain points include productionizing first-party models and balancing AI research against shipping features—a tension common in AI-forward startups still defining their competitive moat.

Tech Stack 39 technologies

Core StackApache Spark AWS Docker Kubernetes Python PyTorch TensorFlow TypeScript React PostgreSQL Redis n8n Terraform Postman Adobe Premiere Pro Apache Flume Dask GCP Azure C/C++ CUDA REST API Cargo Meta Google Segment curl Final Cut Pro Logic Pro Pro Tools+8 more

What Descript Is Building

◆Challenges

Solving complex technical problems
Building legal function from scratch
Integrating ai models into product
Balancing research investment vs feature work
Productionizing 1p models
Expanding to b2b
Scaling pipeline generation
Scaling for larger teams
Scaling for complex workflows
Integrating third-party ai models

▲Active Projects

Third-party model integrations
Ai/ml-powered projects
Model inference optimization
Timeline editing and effects
Ai audio integration
Team media management
Ai research and enablement roadmap
Mlops infrastructure
Model evaluation framework
Demand generation strategy

Hiring Activity

Accelerating20 roles · 20 in 30d

Department

Engineering

Product

Finance

Marketing

Sales

Design

Legal

Support

Seniority

Senior

Director

Junior

Manager

Mid

Company intelligence

Find more companies like Descript by tech stack, pain points and active projects

Get started free

About Descript

Descript makes video, podcast, and social-clip creation accessible to non-professionals by letting users edit video as they would edit text, backed by AI agents that handle design, layout, and effects. The platform serves over 7 million creators and small-to-medium businesses globally. Operationally, the company is engineering-heavy (7 roles), with strong product (4 roles) and emerging commercial functions (sales, marketing, finance each 2 roles), reflecting a shift from creator-focused toward team collaboration and B2B expansion. Headquartered in San Francisco, the company is backed by cloud infrastructure across AWS, GCP, and Azure.

HeadquartersSan Francisco, CA

Company Size51–200 employees

Founded2017

Hiring MarketsUnited States

Frequently Asked Questions

What technology does Descript use for video AI and editing?

Descript runs PyTorch and TensorFlow for model inference, CUDA for GPU compute, and Apache Spark and Dask for data pipelines. The product uses TypeScript and React on the front end, PostgreSQL and Redis for persistence, and Docker and Kubernetes for deployment.

What is Descript working on?

Active projects include third-party model integrations, AI audio synthesis, timeline editing enhancements, MLOps infrastructure, and a demand generation strategy. Pain points reveal tension between AI research roadmap and feature velocity, plus challenges productionizing first-party models and expanding into B2B team workflows.