Lead AI Engineer (LLM & Agents)

• Production LLM/Agent experience at scale. You've built LLM or Agent features used by 10,000+ active users, and you've maintained and improved them after launch.

• Strong backend engineering fundamentals. You're comfortable owning production code, APIs, data flows, and incident troubleshooting. Experience designing and implement backend services in Golang and integrate with LLMs and AI systems.

• Real evaluation discipline. You've built evaluation sets, regression suites, and production monitoring, and you can explain how you confirmed an improvement was real rather than random fluctuation.

• Ownership mindset. You make decisions, surface risks early, and don't consider a feature done until it's verified in production.

• Web3 experience. Preferably 1 year of experience in a crypto, DeFi, or Web3 company (wallet, exchange, L1/L2 protocol, or DeFi platform strongly preferred)

We're looking for a Lead AI Engineer to own the backend and Agent systems behind our LLM-powered products. The role is end to end: you'll build the Agent workflows, define how we measure quality, keep them reliable in production, and shape where the product goes next.

This role is for someone who has shipped LLM features to real users, stayed with them after launch, and can prove with data that their changes made things better.

• Design, build, and maintain multi-turn conversation systems, including context management, tool calling, knowledge retrieval (RAG), state management, and human handoff.

• Define success criteria with Product, Design, and QA leads.

• Turn production failures into evaluation samples and regression tests.

• Measure the impact and side effects of every change before and after release.

• Isolate the root cause of failures: model judgment, knowledge gaps, abnormal tool responses, backend instability, or conflicting rules.

• Apply the right fix for each cause rather than defaulting to prompt tweaks.

• Handle timeouts, retries, graceful degradation, async jobs, and third-party dependency failures.

• Make sure every issue is traceable through logs, every change is versioned, and every failure has a recovery path.

• Give clear, reasoned positions on requirements, covering feasibility, cost, risk, and alternatives.

• Propose better approaches and break work into phased, shippable deliverables.

• Share progress, blockers, ETAs, and next checkpoints without needing to be chased.

• After launch, track real-world impact and leave clear documentation and handover materials.