AI engineer for agent systems

Building the entertainment studio of the future. Looking for AI filmmakers, data analyst, product manager, producers. Join us!

• Agent systems. Fleets of agents that plan, create, critique, and repair autonomously, and stay reliable when nothing about them is deterministic.
• Generative media at scale. Getting consistent, high-quality output from frontier models, hour after hour.
• Behavioural intelligence. Understanding what drives attention and desire, and turning real-world response into better decisions about what to make next.
• Distributed infrastructure. Systems that scale with demand and survive the failures of everything they depend on.
• Consumer product. The owned platform, where audiences engage and convert.

• A senior or staff-level engineer, or a founder or early engineer who has carried a production system on your back. You are still deeply hands-on.
• Strong in Python and comfortable in TypeScript/React.
• Someone who has built distributed or high-throughput systems and debugged them in production.
• Already building with LLMs beyond the demo stage, in production, with real evaluation.
• A heavy, opinionated user of AI coding agents who already ships multiples of what you shipped two years ago.
• Fast, but with a very high quality bar. You can tell a real fix from a patch that hides the symptom.
• A clear writer. Our plans, audits, and postmortems are written documents.
• Будет плюсом: Generative media, video pipelines, or film/TV experience; Growth engineering, ad tech, or experimentation systems; Consumer apps at scale; Eval design for LLM systems; Having founded something, or having been one of the first engineers somewhere

• Extreme ownership. You own an area from design to production to the 2 a.m. incident. Nobody hands you specs. You find the problem, write the plan, ship it, and measure it.
• Agent-native engineering. Most code here is written by humans directing coding agents (Claude Code, Cursor, Codex). The skill that matters is judgment: writing a precise plan, running agents in parallel, reviewing their output hard, and keeping the codebase small and clean. Our standard is that simple, correct code beats clever code.
• Production is the lab. We learn from real-world results, not demos.
• Outcomes are the metric. We judge work by measurable results, not by how impressive it looks.