index

We need the person who makes the product feel calm when usage spikes. You will own the systems that keep conversations fast, character state consistent, and infrastructure costs legible.

Why this role exists

janitor is building interactive AI entertainment for millions of people. That means normal web scale plus unusually heavy inference paths, realtime sessions, large stores of creator and character data, ranking systems, safety pipelines, and bursty fan behavior. The infra has to absorb all of it without becoming a science project.

What you’ll do

• Design and operate Kubernetes, Linux, networking, observability, deploy, and incident workflows.
• Build GPU and inference orchestration that keeps latency predictable and spend visible.
• Scale Postgres, caches, queues, object storage, and background workers without hiding bottlenecks.
• Improve reliability for realtime chat, creator tools, search, moderation, and recommendation pipelines.
• Remove manual ops work by turning repeated fixes into simple internal systems.

What strong looks like

• You have owned production infrastructure with real user pressure, not just maintained Terraform.
• You know when Kubernetes helps and when it is masking a bad boundary.
• You can debug from packet to query plan to deploy trace.
• You write clean backend code when infra work needs product-facing glue.
• You communicate tradeoffs clearly during incidents and architecture decisions.

Nice to have

Experience with high-throughput inference, GPU scheduling, bare metal, realtime systems, or ML platform work.

How to apply

Email [email protected] with a short note and something you have built or operated under pressure.