Skills & Experience

• AI-native in how you build: you use agentic coding tools (Claude Code, Cursor, Codex, or similar) as part of your daily workflow, and you pick up new best practices quickly and bring them to the team.
• 4+ years shipping production web apps across frontend and backend.
• Strong TypeScript and React, plus a Python or Node.js backend and PostgreSQL.
• Has shipped at least one LLM-backed feature, including streaming output to a UI.
• Understands how agents behave in practice: latency, cost, tool calls, and imperfect outputs.
• Cloud deployment and CI/CD experience.

Nice to have

• Experience with AI app tooling such as the Vercel AI SDK, LangGraph's frontend integrations, or building on MCP.
• Familiarity with LLM observability platforms (Langfuse, LangSmith, Arize Phoenix, Braintrust) and OpenTelemetry.
• Data visualization skills (D3, Recharts, or similar) for dashboards and eval results.
• Experience in financial services or other regulated environments with strict access control and audit needs.
• Design chops: comfortable prototyping in Figma or directly in code.

Responsibilities

Build the product experience around agents

• Build web applications where users work alongside agents: chat and copilot interfaces, task-based workflows, and embedded AI features in existing tools.
• Design UX for non-deterministic systems: streaming responses, progress and step visibility, citations and source previews, confidence cues, and graceful failure states.
• Build human-in-the-loop flows: approvals, edits, overrides, and feedback capture (thumbs, corrections) that feed back into evals.
• Handle long-running and async agent tasks with background jobs, notifications, and resumable sessions.

Own the application backend

• Design and build APIs and services that sit between the frontend and the agent layer, including auth, permissions, rate limiting, and multi-tenancy.
• Implement streaming transports (SSE, WebSockets) and queue-based orchestration for agent runs.
• Model data for conversations, runs, tool calls, user feedback, and documents; manage storage across relational, vector, and object stores.
• Expose internal systems and data as clean tools the agents can call, including MCP servers where useful.

Ship and operate

• Own CI/CD, testing, and deployment for the application layer.
• Instrument the app end to end so a user action can be traced through the UI, API, and agent run.
• Guard against AI-specific risks in the app layer: prompt injection via user input, data leakage across tenants, and access-control gaps in tool calls.
• Partner with design, product, and business users to scope features and iterate fast on real usage.