Senior SDE - AI Platform Engineer

What you'll do

• Design, build, and ship LLM-powered features and bounded autonomous agents with observability, latency budgets, and fallback behaviour
• Architect RAG pipelines — chunking, embedding selection, retrieval, and re-ranking — for internal knowledge bases and product surfaces
• Build multi-step agentic workflows (LangGraph, LangChain, CrewAI, Claude Agent SDK) with scoped tool sets, structured outputs, and auditable side-effects
• Standardise AI-assisted engineering workflows — Claude Code, Cursor, GitHub Copilot, Augment Code — driving measurable feature delivery improvements
• Author CLAUDE.md / AGENTS.md context files, shared system prompts, slash commands, and MCP integrations for the engineering team
• Embed AI agents into code-review and test-generation loops with guardrails ensuring AI output meets engineering quality standards
• Partner with Product, Engineering, Data, and Operations to identify and scope high-leverage AI automation opportunities
• Run structured evals and prompt experiments; maintain harnesses that track model quality and alert on regressions after upgrades
• Document AI system designs, agent architectures, prompt libraries, and runbooks so the team can maintain and extend AI features independently
• Instrument LLM features with structured logging, cost tracking, and latency monitoring; own SLOs and incident response for AI-driven workflows
• Apply output validation (Zod / JSON Schema) and human-in-the-loop checkpoints to manage hallucination risk in production

Requirements

• 3+ years building and shipping LLM-powered features or AI agents in production
• Working knowledge of Claude (Sonnet / Opus), GPT-4o, and/or open-source models (Llama, Hugging Face ecosystem)
• Proficiency with agentic frameworks: LangGraph, LangChain, LlamaIndex, or CrewAI; experience designing bounded, auditable agent workflows
• Hands-on RAG experience: chunking, embeddings, vector stores (Pinecone, ChromaDB, pgvector), retrieval strategies, and re-ranking
• Strong prompt and context engineering: system prompt design, structured output enforcement (JSON Schema / Zod), and prompt caching
• Experience with Claude Code — context files (CLAUDE.md), slash commands, and agent guardrails
• Experience with Cursor — AI-assisted coding and codebase-aware refactoring workflows
• Experience with GitHub Copilot and/or Augment Code in daily development and code-review cycles
• Experience with MCP (Model Context Protocol) — configuring or authoring MCP servers for agent context management
• Strong TypeScript / JavaScript and Node.js skills; Python for ML tooling is a plus
• Experience with React or equivalent front-end framework; ability to build streaming, real-time AI-powered UIs
• Experience with AWS or equivalent cloud (Lambda, API Gateway, S3, DynamoDB); REST, WebSockets, SSE, OAuth 2.0
• Experience with observability tooling applied to LLM systems: structured logging, cost tracking, latency dashboards

Tech stack