Technical Program Manager, AI Performance

Figma Figma · Enterprise · Canada +1 · Engineering

Figma is seeking an AI-Native Performance TPM to manage performance for both flagship products and AI features, focusing on prevention, diagnostics, and safe rollout across various platforms. The role requires expertise in performance testing, observability, and incident response, with a focus on AI features like model inference latency, throughput, cost, and reliability.

What you'd actually do

  1. This is a specialist Platform TPM owning horizontal, high-visibility performance programs that span:
  2. Flagship product performance (load times, FPS, memory) and AI features (model inference latency, throughput, cost, and reliability)
  3. Desktop, browser, native mobile (React Native / WebView), and WASM contexts
  4. Cross-org programs: observability/telemetry, regression prevention (performance CI), xfn performance forum, SEV mitigation, safe rollout of AI capabilities, and CE Planning (customer engineering / enterprise readiness)

Skills

Required

  • 5+ years in performance engineering, performance TPM, platform TPM, or SRE with hands-on experience shipping performance programs for SaaS products
  • Demonstrated experience with load, stress, performance or scalability testing and new-build comparisons
  • Deep familiarity with web performance (FCP, LCP), rendering/FPS, WASM memory, mobile profiling (Xcode Instruments, Android Profiler)
  • CI/CD and test automation experience: integrating performance tests into pipelines
  • Distributed systems, cloud infra (AWS), autoscaling, and capacity planning
  • Hands-on prior work with LLMs, prompt engineering, or agent platforms

Nice to have

  • Experience with MLops or LLMs: inference latency tradeoffs, batching, caching, etc
  • Observability stack: Datadog / Sentry / RUM
  • Background as a performance engineer (not just TPM) with code or test harness contributions.
  • Experience with large enterprise Planning / customer load tests.
  • Familiarity with real-time collaborative systems and GPU/WASM performance constraints.

What the JD emphasized

  • performance prevention
  • diagnostics
  • safe rollout
  • model inference latency
  • throughput
  • cost
  • reliability
  • performance testing
  • observability
  • incident response
  • LLMs
  • prompt engineering
  • agent platforms

Other signals

  • performance prevention
  • diagnostics
  • safe rollout
  • model inference latency
  • throughput
  • cost
  • reliability