AI Agent Overlord
Here you go,
-----
**AI AGENT OVERLORD**
*OneReach.ai | Berkeley, CA (or wherever the signal reaches)*
**Reports to:** The CEO, technically. The emergent swarm intelligence, spiritually.
**Compensation:** Competitive. Plus the intangible reward of being one of the few humans who actually understands what’s happening.
-----
**About the Role**
We’re looking for someone to command our growing army of AI agents. Not in a Skynet way. More like a really chill conductor who happens to be orchestrating thousands of autonomous digital entities across healthcare, government, and enterprise systems that millions of people depend on daily.
No pressure.
You’ll be building ecosystems of agents that learn from each other, teach each other, and occasionally surprise us in ways that make us nervously laugh and then update our monitoring dashboards. You’ll wield neurosymbolic architectures like a wizard who actually took the time to understand the physics behind the spells.
-----
**What You’ll Actually Do**
Design and deploy agent hierarchies that would make org chart consultants weep with confusion and joy. Build meta-learning systems where agents improve other agents, creating a virtuous cycle we’re choosing to call “helpful” rather than “concerning.” Architect knowledge systems that remember everything important and gracefully forget everything that doesn’t matter, which is somehow harder than it sounds. Translate between the world of probabilistic language models and the world of people who need things to work correctly 100% of the time in FedRAMP-certified environments. Occasionally explain to non-technical stakeholders why the agent did *that* and whether it will do it again.
-----
**You Might Be a Fit If**
You’ve read the neurosymbolic papers and thought “yes, but what if we actually shipped this.” You understand that LLMs are very impressive and also kind of like a brilliant intern who confidently makes things up. You’ve built agent systems that ran in production, not just in notebooks that worked once. You find the phrase “tool use orchestration” genuinely exciting. You can hold the tension between “move fast” and “this runs Medicare enrollment.”
-----
**Bonus Points**
You’ve made agents collaborate without them descending into infinite loops of polite mutual delegation. You’ve debugged an agent that was technically doing what you asked but in a way that revealed you asked the wrong thing. You have opinions about memory architectures and those opinions have changed at least twice.
-----
**Why This Job Exists**
We built the only FedRAMP-certified AI agent runtime. We’re deploying agents in environments where “oops” isn’t really an option. We need someone who can make these systems smarter, more autonomous, and more trustworthy simultaneously—which is the whole game, really.
-----
**Interview Process**
A conversation about what you’ve built. A technical deep-dive that’s more “show us how you think” than “whiteboard this algorithm.” A vibe check to make sure you’ll thrive in a place where the CEO sends Claude transcripts at odd hours with subject lines like “what if agents could…”
-----
*OneReach.ai is an equal opportunity employer. We welcome all humans to apply. Agents, please use the other form.*
-----