Staff Site Reliability Engineer - Dev Platform
Kraken, Brazil +5
The role
- LLM traffic doesn't behave like normal API traffic. Requests are long-running, streaming and expensive. One prompt can cost a few cents or a few dollars. One team's burst can burn another team's budget. Providers change their rate limits, failure modes and pricing without warning. Running this well is a problem few companies have solved, and it's the job.
- Own the reliability of the platform engineers depend on: the AI Gateway, the agent runtime and the connections to internal tools. Set the SLOs, own the error budgets and decide what happens when they're spent.