Senior Software Engineer - Kimchi

Why Kimchi? Kimchi is the AI platform inside CAST AI. We started by helping companies run LLMs on their own Kubernetes clusters and now we're providing a managed variant of those same capabilities. Our Infrastructure today Multi-model inference (MiniMax, Kimi, GLM-5, Nemotron, DeepSeek) with intelligent routing, an OpenAI-compatible API and deployment ranging from our GPUs to your own VPC. The inference layer is the foundation and the API is what sits in front of it as the primary channel for broadly and reliably distributing our AI services and powers our own Kimchi harness. We are hiring across multiple teams! As a Senior Software Engineer, you will have the opportunity to work on different key features of our product. All of these are high-agency roles across multiple parts of the tech stack that minimize process friction that would otherwise prevent you from shipping. In every team you will own features end-to-end: design, implementation, testing, production rollout. Most projects ship in 1-4 weeks. You'll work directly with product and other engineering teams on problems that don't have textbook solutions. We are currently hiring Senior Software Engineers for the following teams: API Platform Owns the inference API responsible for delivering our AI services across the world while making sure it's reliable and capable to scale in tandem with our company's growing ambitions, as well as our analytical platform, billing and role-based controls that enable our users to monitor and control their usage with ease - be it as a solo developer or a large enterprise. You'll own our infrastructure, datastores, analytics, observability and CI/CD pipeline. Responsibilities: Develop with observability in mind, identify bottlenecks and optimize for performance. When p99 latency climbs, you find the cause through query profiles and flame graphs instead of raising the alert threshold. Design the datastores and distributed systems behind the inference API, and keep them reliable as usage scales from a solo developer to a large enterprise. Build the billing, usage metering, and role-based controls that let users monitor and govern their own consumption. Catch correctness problems where a wrong number costs you trust. Harness OpenAI and Anthropic ship models. They also ship one harness each – the scaffolding that turns a raw model into something that can plan, execute, recover, and complete work. We ship a different kind of harness: one built for cost-conscious, long-horizon autonomy, running on inference infrastructure we control end-to-end. A decent model with a great harness beats a great model with a bad harness. We've watched this play out. The gap between what today's models can do and what you see them doing is largely a harness gap – and that gap is where we operate. Responsibilities: Architect planner/executor/evaluator pipelines – planning with a reasoning model, execution with a fast one, evaluation with a third. No self-verification. Manage agent memory and context – state persistence across sessions, context compaction, tool-call offloading Own the harness surface - TUI, MCP integrations, telemetry. Work directly with users - including our own colleagues - to identify, understand and fix pain points Agent Platform The Agent Platform squad builds the infrastructure that lets teams run fleets of AI agents securely, whether in the cloud or self-hosted, so work can move confidently from a developer's laptop to production environments. The platform enforces least-privilege access to resources, requires human approval before agents can touch anything critical, and logs every action agents take for full auditability. The goal is to give engineering and operations teams the confidence to scale up agent usage without sacrificing security, control, or visibility into what their agents are actually doing. Responsibilities: Design and build distributed systems that run agent fleets at scale across cloud and self-hosted environments, with sec…