Description
Summary We're looking for a senior engineer who's actually built and shipped agentic AI systems in production — not someone who's read about them. The work: designing and building multi-agent workflows that plan, use tools, and complete multi-step tasks reliably. Think autonomous research agents, workflow automation agents, or AI systems that need to reason across multiple steps and recover from failure gracefully. What you'll be doing: Architecting agent orchestration (planning, tool-calling, memory, state management) Building and tuning RAG pipelines that actually retrieve the right context Working with frameworks like LangGraph, CrewAI, AutoGen, or a custom orchestration layer — whatever fits the problem Integrating LLM APIs (OpenAI, Anthropic, open-source models) with production backend systems Handling the unglamorous but critical stuff: error handling, retries, evaluation/observability, cost control on token usage Collaborating directly with our team — this isn't a "here's a spec, go build it in isolation" gig Must-haves: 5+ years production ML/AI engineering, with hands-on agentic AI or LLM orchestration work in the last year Strong Python; comfortable with async patterns and API integration Real experience with at least one agent framework (LangChain/LangGraph, CrewAI, AutoGen, or equivalent custom system) Can explain a system you built — not just "I used GPT-4" but the actual architecture decisions and trade-offs Solid communication in English, async-friendly, can work with limited overlap in working hours Nice-to-haves: Experience with vector databases (Pinecone, Weaviate, pgvector) Familiarity with evaluation frameworks for LLM outputs Prior fintech or compliance-heavy domain experience How to apply: Skip the generic cover letter. In your first line, tell me one agentic AI system you've built and what specifically made it hard to get right. Applications that don't do this get skipped.