OmniNode recruits ML engineers for agent infra

This looks less like a chatbot startup and more like an attempt to turn agent workflows into something you can actually operate in production. The bet is sensible: the hard problem in agent systems is not prompting, it is control, replay, and proof that work really happened.

• –Contract-driven delegation is the right abstraction if they want predictable behavior across agents and machines
• –Kafka-backed event sourcing and replayable state suggest they’re treating debugging and auditability as first-class, which most agent stacks still do not
• –Multi-model routing across local and cloud inference is pragmatic, but it raises the bar on routing logic, latency, and failure handling
• –Evidence-backed completion is the strongest part of the pitch because it moves “done” from model assertion to durable verification
• –The hiring message implies this is still early and infra-heavy, so the interesting signal is the systems mindset more than a polished product surface