Tavus Tech Stack

AI avatar platform for real-time human simulation and conversational agents

Software Development San Francisco, CA 51–200 employees Founded 2020 Privately Held

Tavus is a research-heavy organization building generative video and conversational AI models—the stack is PyTorch + Node.js + React Native, with heavy lifting on GPU infrastructure and RAG systems. The hiring mix (research nearly matches engineering, senior roles dominate) reflects a company still in productization phase, moving research-lab outputs toward production systems. Current friction points—data labeling, video structuring, preventing production issues, faster customer integration—signal the tension between cutting-edge model development and operational maturity at scale.

Tech Stack 15 technologies

Core StackReact Native React RAG Next.js Node.js Firebase Supabase Vercel TypeScript Python PyTorch AWS Kubernetes GCP AWS EKS

What Tavus Is Building

◆Challenges

Preventing production issues
Debugging api requests
Unblocking customers
High-quality data procurement
Video data structuring
Labeling automation
Debugging integrations
Unblocking adoption
Customers integrate faster
Scaling team

▲Active Projects

Generative video models
Productionizing research
Gpu infrastructure extension
Rag and memories
Production system integration of multimodal models
Llm and vlm research for conversational avatars
Agentic systems and features (rag, memories, mcp, async task management, job scheduling)
Developer portal
Neural avatars
Scalable motion language

Hiring Activity

Decelerating45 roles · 10 in 30d

Department

Engineering

Research

Design

Data

Sales

Support

Marketing

Seniority

Senior

Mid

Director

Manager

Company intelligence

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

Get started free

About Tavus

Tavus builds AI avatar technology for real-time conversational interactions. The product synthesizes audio, video, and language models to create simulated human agents capable of natural dialogue in multiple languages. The company operates as a research lab with parallel engineering effort; the active project list (neural avatars, scalable motion language, agentic systems, multimodal model integration) shows focus on both core model research and production infrastructure. Founded in 2020 and based in San Francisco with 51–200 employees, the organization is hiring across engineering and research roles in the United States and United Kingdom.

HeadquartersSan Francisco, CA

Company Size51–200 employees

Founded2020

Hiring MarketsUnited States, United Kingdom

Frequently Asked Questions

What technology does Tavus use to build AI avatars?

Tavus builds on PyTorch for model development, Node.js and React Native for application layers, and RAG (retrieval-augmented generation) for memory and contextual response. Infrastructure runs on AWS/GCP with Kubernetes orchestration and GPU extensions for inference.

What is Tavus working on?

Tavus is developing generative video models, conversational avatar systems, scalable motion synthesis, and agentic systems (task scheduling, memory, async job handling). Current focus includes productionizing research outputs and extending GPU infrastructure for inference scaling.