SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
edisyl builds AI solutions that transform messy institutional data into decisions, workflows, and outcomes. The company emerged from 8 years of blockchain data infrastructure work (20+ chains, 700M+ resolved wallets) and now applies that capability to enterprises facing the challenge of making data work at scale without armies of analysts. They have active deployments with financial institutions and proven architecture, with inbound interest from firms needing their solutions.
As AI Agent Architect, you own the architecture that makes agent fleets reliable: the harness, tooling, orchestration patterns, and semantic layers that keep outputs grounded in organizational context. You work on Forge (agent framework), Lattice (fleet orchestration), and Stratum (semantic intelligence) — systems that production deployments run on. This is a Staff-level role where you define architecture, set standards, and make principled decisions independently.
Key responsibilities include: designing and building architecture for AI agent workflows (planning loops, tool use, memory, retrieval, human-in-the-loop checkpoints); evaluating, integrating, and fine-tuning foundation models and LLM APIs for enterprise use cases; defining standards for agent reliability, observability, and failure modes in production; collaborating with Forward-Deployed Engineers to translate client learnings into reusable platform components; building internal tooling and eval harnesses to assess agent quality and hallucination rates; and making documented architectural decisions while staying current with the ecosystem.
Success in year one means shipping meaningful improvements to core systems running in production, establishing the eval framework the team uses, enabling Forward-Deployed Engineers to focus on client problems rather than infrastructure workarounds, and making architectural decisions that the company builds on for years.
You bring 6–10 years building production AI or data systems (not prototypes), deep hands-on experience with multi-agent architectures, strong Python skills and familiarity with agent frameworks (LangChain, LlamaIndex, AutoGen, or equivalent), practical RAG and vector database experience in production, and enterprise LLM deployment experience (data security, access controls, audit logging). Critically, you have LLM failure mode literacy, production instincts, strong opinions on agent design, and systems thinking focused on failure modes first. Bonus experience includes ML research exposure, regulated industry deployments (financial services, insurance, healthcare), and blockchain data infrastructure familiarity.
