Frontend Engineer — Geospatial Visualization

Build polished, performance-conscious React/TypeScript UI for map-driven workflows: 3D entity rendering, real-time status, modal-driven dispatch, complex form ergonomics. Own the UX of multi-step operator flows end-to-end — visual hierarchy, loading/error states, keyboard, accessibility, edge cases. Collaborate on AI-driven features: integrating Claude Agent SDK tool flows, designing prompts and MCP tools, surfacing agent actions and audit logs in the UI. Occasional Python work on FastAPI services when frontend changes need backend hooks. Must-haves 5+ years shipping production React apps with TypeScript. Heavy day-to-day AI tooling — Cursor / Claude / agentic workflows. You move fast with AI, refactor with it, and reason about prompts and tool design (not just autocompletes). Strong UI/UX intuition. You see a Figma or sketch and can ship something that feels right; you push back when the design hurts the user. Works independently and takes initiative — you scope your own work, surface unknowns early, and ship without hand-holding. Detail-oriented — you sweat the loading states, the empty states, the off-by-ones, and the cases everyone else forgot about. Comfort with styled-components, modern hooks, Redux (or equivalent), and careful state design in long-lived single-page apps. Nice-to-haves CesiumJS, Leaflet, Mapbox, or Three.js — anything map / 3D. Python (FastAPI + Postgres) for working across the stack. Experience designing or consuming MCP servers, or building tool-using agents. Real-time / streaming UI experience (SSE, WebSockets, telemetry). A few hard truths before you apply AI is a tool, not a substitute for engineering. We use Cursor and Claude every day, but you still need to understand the code AI produces, debug it when it breaks, and explain trade-offs and decisions when asked. If AI is your only path to a working PR — and you can't defend the PR once it lands — this won't be a fit. You'll inherit complexity. This is a real codebase with history. You'll spend time reading before you write. The bar for "done" is high. "It works on my machine" doesn't ship. We care about edge cases, accessibility, regressions, and the ten things the user might do that you didn't think of. Generic AI-generated cover letters get instant rejection. We can tell. So can the next reviewer.