Camera Isp Software Engineer, Consumer Devices

OpenAI OpenAI · AI Frontier · San Francisco, CA · Consumer Products

This role focuses on developing and tuning camera Image Signal Processing (ISP) pipelines for consumer devices, ensuring high-quality imaging under real-world constraints. It involves end-to-end ownership from prototype to production, collaborating with hardware and firmware teams, and building robust evaluation workflows. While the devices use AI, the core craft of this role is in traditional ISP software engineering, not direct AI/ML model development.

What you'd actually do

  1. Own end-to-end ISP tuning and bring-up, from early prototypes through production hardware.
  2. Tune core 3A and image-quality blocks including AE, AWB, AF, noise reduction (spatial and temporal), sharpening, tone mapping, color correction, HDR/WDR, flicker mitigation, lens shading, defect pixel correction, and related ISP stages.
  3. Build repeatable capture and evaluation workflows, combining controlled lab sweeps with real-world validation.
  4. Validate robustness across sensor, module, and manufacturing variation with clear, defensible success criteria.
  5. Deliver production-grade tuning artifacts, including versioned tuning packs, parameter manifests, change logs, and curated RAW and processed datasets with supporting documentation.

Skills

Required

  • ISP tuning for embedded or mobile camera products
  • 3A fundamentals (AE/AWB/AF)
  • noise modeling
  • sharpening and detail trade-offs
  • tone mapping
  • color pipelines
  • data-driven tuning loops
  • repeatable capture
  • rigorous evaluation
  • disciplined versioning of tuning artifacts
  • debugging end-to-end camera pipelines on real hardware
  • RAW capture
  • processed outputs
  • hardware-accelerated paths
  • cross-functional collaboration
  • translating system-level constraints (power, latency, motion, reliability) into tuning decisions and validation plans

Nice to have

  • common mobile or embedded SoC camera stacks and vendor frameworks
  • tuning imaging pipelines explicitly for downstream machine-learning consumers
  • iteration driven by model feedback
  • designing motion-aware temporal imaging strategies

What the JD emphasized

  • end-to-end ISP bring-up and tuning
  • production hardware
  • real-world constraints
  • shipping production-grade imaging systems
  • deep technical ownership
  • end-to-end ISP tuning
  • production hardware
  • production-grade tuning artifacts