About the roleWe're hiring a backend engineer to build and scale the infrastructure behind FLORA's multi-modal AI engine.
You'll work directly with the founders. The core problem: creatives run heavy, multi-step workflows on the canvas in real time, and the backend infra has to keep up. Your main job is a centralized execution engine that manages every model and tool we plug in.
What you'll doBuild the architecture that keeps the canvas fast and reliable as we scale to more generative models and tools.
- Own and design the execution engine. One system that routes every generative tool, external model API, and canvas transformation through a standard path.
- Speed up the canvas and cut data overhead.Cut database query overhead. Move tool configs and execution state into caching and in-memory stores so the canvas stays fast even when a workflow spans many nodes.
- Handle failure well. Generative APIs time out, hit rate limits, and choke on big payloads. Build retries, circuit breakers, and fallbacks into the routing layer so one failed model doesn't kill the user's session.
- Scale the multi-modal pipeline to millions of users. Work with product and applied AI to move, process, and store heavy payloads (text, high-res images, video) without bottlenecks.
Requirements- 3+ years on backend infrastructure, platform engineering, or distributed systems.
- Hands-on experience building execution layers, execution graphs, state machines, or routing services.
- A track record of finding database bottlenecks and fixing them: caching, query tuning, schema redesigns.
- Comfort with multi-modal infrastructure: managing state and streaming for text, image, and video.
- You care how backend latency and reliability show up in what the user can actually do on the canvas.
- TypeScript familiarity and comfort in a serverless/edge runtime world
Bonus- Event-driven architectures, durable-workflow/state-machine engines, or graph execution engines.
- Backends for node-based, real-time, or collaborative (CRDT/multiplayer) interfaces.
- Familiarity with our stack - Convex, Liveblocks, Vercel Workflow, sandboxed code execution - or experience standing up an internal model/config registry.
