SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Relay is a cross-chain payments protocol providing instant, low-cost bridging and transactions across 85+ chains, serving 5+ million users with 59+ million completed transactions and $6+ billion in volume. The company's mission is to make cross-chain transactions as fast, cheap, and reliable as online payments, abstracting blockchain complexity from end users.
You will join the Developer Experience team as a Senior AI Agent Engineer, owning Relay's internal Slack-native AI agent. This is a hands-on builder role where you are measured by what the agent accomplishes accurately and safely. You will cover US business hours with dependable overlap with the teams whose workflows you're automating.
Key responsibilities include:
- Own and continuously improve the agent's backlog, prioritizing by impact, frequency, and risk
- Review real misses weekly and convert them into regression evaluations, fixed tools, or improved guidance
- Transform expert workflows into agent skills, decision logic, and escalation paths
- Build knowledge systems that help agents find, connect, and reason over Relay's structured and unstructured information with clear provenance and freshness tracking
- Improve retrieval, orchestration, and verification; own regressions through resolution
- Measure task success, correction rate, latency, and cost; report metrics regularly
- Integrate agents with internal systems through APIs and MCP tools safely
- Choose between deterministic code, simple automation, or agent behavior based on merits
- Treat permissions, side effects, human approval, and failure recovery as product features
You are a product-minded engineer who ships AI systems into real workflows. You care deeply about correctness, permissions, and failure handling alongside model capability. You read traces before changing prompts and prefer adding linters, types, or tests over prompt tuning. You know when conventional software outperforms agents. You can extract decision logic from domain experts and encode it into testable behavior. You're comfortable being the first dedicated hire for a new function.
Requirements:
Must-have:
- 5+ years in software engineering, developer tooling, or platform engineering
- Production experience with LLM agents (tool use, retrieval, structured outputs, orchestration)
- Experience building search, retrieval, or knowledge systems across structured and unstructured information
- Strong TypeScript or Python with API integration experience
- Experience designing evals or automated tests for nondeterministic software
- Familiarity with permissions, secrets, auditability, and safe handling of side effects
- Strong written communication; ability to turn expert workflows into precise specs
Nice to have:
- Experience with MCP or comparable tool and context protocols
- Experience with Slack apps, support automation, or internal platforms
- Experience with entity resolution, knowledge graphs, ontology/taxonomy design, ranking, or semantic enrichment
- Experience in crypto, payments, or other high-stakes domains
