Vision Systems Architect

Role overview

Responsibilities

• Define system architecture from sensor and ISP through CV/ML to the application layer; translate product and UX requirements into scalable solutions.
• Set image-quality strategy and architect reusable, modular CV/ML frameworks across product SKUs.
• Balance performance, power, cost, and thermal limits; define AI model lifecycle strategy from training to OTA deployment.
• Architect CPU/GPU/NPU/DSP utilization and set KPIs for latency, FPS, power, memory, boot time, and robustness.
• Select inference engines, frameworks, and hardware platforms; guide SoC bring-up and vision-pipeline performance validation.
• Establish coding standards, modularization, and maintainability; identify technical risks and mitigations.
• Collaborate with hardware, ISP, Android/Linux platform, and product teams; mentor senior engineers and provide architectural governance.
• Contribute to the long-term vision/AI roadmap across product generations.

Mandatory qualifications

• Bachelor's or master's degree in Computer Science, Electrical Engineering, or a related field; PhD preferred.
• 10–15+ years of experience in imaging, vision, or perception systems.
• Proven experience architecting and shipping multiple consumer products with vision/AI.
• Deep imaging-pipeline knowledge, including sensor behavior and ISP tuning impact; strong embedded systems architecture experience.
• Expert-level C++ and system-level optimization knowledge; strong CV/ML integration expertise.
• Experience deploying optimized ML models on consumer SoCs and with heterogeneous compute architectures.
• Experience defining system-level KPIs and validation strategies; strong cross-functional leadership and communication skills.

Nice-to-have experience

• Computational photography; AR/VR or advanced camera features; silicon bring-up and hardware/software co-design.
• Large-scale model deployment and OTA updates; multi-camera or depth-sensing systems; patents or technical publications.
• Architecture leadership across multiple product generations.