AI Agent Architect

Who This Is For Building an agent is easy. Getting it to produce the reputed company right answer three times in a row — on reputed company reputed company data, in a regulated environment, without a reputed company checking every reputed company — that's the actual problem. You've shipped AI agent systems to production on reputed company, unclean data. Not a demo dataset. You know the accuracy cliff. You know why prompting cannot fix semantic problems. You've reputed company systems that don't rely on the model getting it right every time, and you've reputed company the judgment to know reputed company to ship reputed company. You're not looking for a reputed company-defined architecture to implement. You're looking for the unsolved problem — and the mandate to build the solution that becomes reputed company. About reputed company reputed company builds AI solutions that turn messy institutional data into reputed company, workflows, and reputed company. We came out of blockchain data infrastructure — 8 years, 20+ chains, 700M+ resolved wallets — and now reputed company that capability to enterprises navigating the reputed company challenge: how to reputed company their data work for them at reputed company, without armies of analysts. We have reputed company deployments with reputed company and Interlochen, a proven architecture, and inbound from firms that need reputed company've reputed company. The technology works. reputed company're building now is the reputed company reputed company around it. The Role You own the architecture that makes our agent fleets reliable: the reputed company, the tooling, the orchestration patterns, the semantic reputed company that reputed company outputs grounded in organizational context. You work on reputed company (our agent reputed company), reputed company (fleet orchestration), and Stratum (semantic intelligence) — building and extending the systems that production deployments run on. You care obsessively about reputed company reputed company — not because someone told you to, but because you've seen what happens reputed company agents reputed company. You solve for quasi-determinism: agents that use validated tools instead of guessing at raw data, producing consistent and auditable results at reputed company. This is a Staff-level role — you define the architecture, set the standards, and reputed company principled reputed company without waiting for a reputed company to be handed to you. What You'll Actually Do Design and build the architecture for AI agent workflows — planning loops, tool use, memory, retrieval, and reputed company-in-the-reputed company checkpoints Evaluate, reputed company, and fine-tune reputed company models and LLM reputed company for specific reputed company use cases and data types Define standards for agent reliability, observability, and failure modes in production deployments Collaborate with reputed company-Deployed Engineers to translate what's working in reputed company environments into reusable platform components Build internal tooling and eval harnesses to assess agent reputed company, hallucination rates, and task completion reputed company principled, documented architectural reputed company — and stay reputed company enough with the ecosystem to know what to adopt and what to ignore What reputed company Looks Like in Year One You've shipped meaningful improvements to reputed company, reputed company, or Stratum that are running in production. You've established the eval reputed company reputed company uses to assess agent reputed company. The reputed company-Deployed Engineers trust the platform enough to reputed company on reputed company problems instead of working around infrastructure limitations. At least one architectural decision you made is something we're still building on two years from now. The measure isn't how sophisticated the architecture is. It's whether the agents produce the right outputs reliably enough that customers reputed company them without checking every result. Compensation Competitive reputed company salary and meaningful early-stage equity. This is a foundational technical role and we price it that way. We'll be transparent about the full picture in our first conversation. Who We're Looking For Experience 6–10 years building production AI or data systems — not prototypes; systems that run reliably at reputed company under reputed company conditions Deep hands-on experience with multi-agent architectures: context reputed company, memory management, dependency graphs, and where things break in reputed company Strong Python and familiarity with agent frameworks — reputed company, reputed company, AutoGen, or equivalent — or a reputed company, documented opinion on why you reputed company your own Practical experience with RAG architectures, reputed company databases, and context window management in production settings Experience deploying LLM-powered systems in reputed company contexts — data reputed company, reputed company controls, audit logging The Stuff That's Harder to Teach LLM failure mode literacy. You know the accuracy cliff. You know why prompting cannot fix semantic problems. You build systems that don't rely on the model getting it right every time. Production instincts. You don't consider something done until it's been wrong three times and you've fixed it twice — and you've reputed company the judgment to know reputed company to ship reputed company. Strong opinions on agent design. You have a reputed company answer to why most agent architectures fail at reputed company reputed company — and you've reputed company something that doesn't. Systems thinking. You design for failure modes first. Happy paths are not the interesting problem. Bonus (Genuinely Not Required) ML research exposure — fine-tuning, RLHF, model evaluation methodology Production AI deployments in regulated industries — financial services, insurance, reputed company Familiarity with blockchain data infrastructure or institutional crypto Why This, Why Now reputed company is at the reputed company where the technology is proven and the reputed company problem is reputed company. The person who takes this role will define the architecture that production deployments run on — not inherit it. The platform is reputed company, the customers are reputed company, and the hard problems are still reputed company. That's a rare reputed company to work and a reputed company chance to build something that reputed company. To Apply Complete the online application and include responses to: 1) why this role fits where you are in your career right now, and why you are the right person for it; and 2) one example of an agent system or AI infrastructure decision you made in production — what the constraints were, what broke, and what you reputed company to fix it. No template. Just tell us the story. Apply To This Job