About Mux

Mux is video for developers and their agents. We handle the hard problems of video, like encoding, streaming, delivery, and monitoring, along with the hard problem of understanding it: turning raw footage into context, structure, and answers. Video is a huge part of people’s lives, and we want to help make it not just playable, but understandable.

We’re committed to building a healthy team that welcomes a diverse range of backgrounds and experiences. We want people who care about our mission, are ready to grow, believe in our values (from Be Human to Turn Customers Into Fans), and want to make the people around them better.

You’ll be joining a tight-knit team with experience at places like Google, YouTube, Twitch, Zencoder, Fastly, and more. Our founders previously started (and sold) Zencoder, an early leader in cloud video technology, and authored Video.js, the biggest HTML5 video player on the web. We organize Demuxed, the premiere conference for video engineers in the world.

We’re backed by top investors like Coatue, Accel, Andreessen Horowitz, and Y Combinator. You’ll get to work with amazing companies: hundreds of startups, plus Patreon, Substack, Synthesia, and HubSpot. Customers large and small love working with us and love our team.

We are building something big together. We’d love to hear from you!

About the Role

Mux makes video easy for developers. Our APIs handle encoding, streaming, playback and analytics for companies like Patreon, Substack and Robinhood.

We're changing how Mux uses its own data. We're not building more dashboards. We're building semantic models that are clean, tested and well documented, and putting AI agents on top of them so anyone at Mux can ask a question in Slack and get a correct answer. Agents on top of bad data just give confident wrong answers faster. So this role is fundamentally a data engineering role: the warehouse, the pipelines and the dbt models have to be right before any of the AI work is worth anything.

As a Senior Data Engineer on the Analytics team, you'll own the Mux data stack end to end, from ingestion through Snowflake and dbt to the semantic layer the agents query. The team is small, so you'll have a broad scope and your work will be visible. You'll also be expected to use AI tools heavily in your own work, and to help decide where they belong in the data stack and where they don't.

This role is a fit if you want to build the foundation an AI-native analytics function runs on, and you'd rather delete a model than add another one.

What You'll Do

Own the Data Platform End to End

Own the Mux data warehouse (Snowflake), the ingestion layer and the transformation layer (dbt): architecture, administration, permissions, CI/CD, cost, ingestion standards and data quality

Build and maintain ingestion from production systems and business tools (Fivetran or equivalent) into the warehouse, and keep it in sync as the product changes

Run the platform day to day: orchestration, deployments, environment management, and incident response when something breaks

Push curated data back out to business systems like Salesforce, NetSuite and HubSpot

Semantic Models and Agents

Build the semantic models and data foundation behind the agents people at Mux use to answer business questions

Help measure whether those agents give correct answers

Data Quality and Reliability

Keep the data accurate, tested and simple: move logic upstream, enforce contracts, and add observability

Pay down tech debt as you go and leave the codebase simpler than you found it

Work AI-Forward

Use coding agents (e.g. Claude Code, Cursor) as a normal part of building, testing and refactoring pipelines and dbt models

Look for places where agents and automation can replace manual data work, and build them

Set standards for how AI tools touch production data: access, review, and what has to be verified by a human

Cross-Functional Partnership

Work with Engineering on how data is produced, changed and retired at the source

Partner with Finance, FinOps, Product and GTM to understand the questions they need answered and decide which models get built first