Full-Stack Engineer

Human Delta is building the continuous-improvement infrastructure for enterprise AI agents, a control plane that governs the knowledge underneath production chatbots and agents and closes the loop between how they perform live and the knowledge they rely on. Despite being a small, sub-10-person seed-stage company ($3M raised from Susa Ventures and Nicole Bischoff), it already holds contracts with household enterprises, T-Mobile, ESPN, Disney, Paramount, across media/entertainment, telecom, and financial services, with advisors from Disney, BCG, and Accenture.

It's rare for a company this early to run real, high-stakes deployments with Fortune-50 brands, and the team is deeply hands-on. Human Delta deliberately doesn't build the agents, it trusts the vendors (Decagon, Sierra, and the like) for that, since agents are already good enough for CX and IT use cases. What those agents actually need is clean, complete, compliant knowledge and a feedback loop from production data. Human Delta identifies what went wrong, corrects the underlying cause, and continuously improves both the knowledge and the agents using it, building the evals and observability so every failure becomes a signal that makes the whole system more reliable.

It sells into regulated industries (customer-support and IT teams) helping them deploy agents safely.

Human Delta is hiring a backend-leaning Full-Stack Engineer to build the infrastructure behind its current platform and next generation of agent products. Most of the hard problems live in the backend, data modeling, queues, execution pipelines, integrations, observability, reliability, and debugging distributed workflows, but you should be comfortable taking a feature through the full stack and shipping the product surface that makes the underlying system useful.

You'll have meaningful ownership over architecture and product decisions, working directly with a small, highly technical team and, when needed, enterprise customers. This is hands-on category-defining work: the platform that decides whether an agent actually got better, running against real Fortune-50 deployments rather than hypothetical demos.