We build AI agents that look and sound human - real face, real voice, real-time conversation. The kind of thing that makes people ask if it's actually live the first time they see it. Getting there means solving latency at a level most infrastructure can't touch, running custom frontier models for lipsync and face generation, and orchestrating all of it live across a globally distributed GPU fabric we own and operate ourselves.
This isn't a role where you maintain what someone else built. The hard problems are still open: dynamic routing and load balancing across thousands of concurrent GPU servers spanning multiple clouds and regions, keeping latency low enough that conversation feels natural, and deployment pipelines that can keep up with a product that's moving fast.
You'd set the technical direction for all of it, working directly with the founders on a 15-person team. No layers between you and the decisions that matter.
If you've worked on distributed systems at scale, care about how things are built and not just that they work, and want real ownership rather than a slice of someone else's architecture — we'd love to hear from you.
Stack: AWS, TypeScript or Go, Terraform/Pulumi/AWS CDK, Docker, serverless architectures.
Looking for: 5+ years hands-on AWS in production, strong IaC, multi-cloud experience, low-latency systems, and startup experience.
We're not able to offer relocation or visa sponsorship. If you're already in Berlin, this is a hybrid role. If you're elsewhere, we're open to fully remote within CET ±2h. Applications outside these criteria won't be progressed.
Apply or read more at https://ojin.ai and please mention you found us on Hacker News in your application. We're also hiring across sales and marketing.