Job Information
Meta AI Specialist - Computational Photography in Zurich, Switzerland
Summary:
We are looking for a strong AI specialist with industry experience in developing and deploying machine learning (ML) computational photography algorithms that run in real-time on mobile or AR/VR platforms.The mission of the Core AI Computational Photography team is to extract the highest imaging performance possible out of the physically-constrained cameras on AR and VR by developing device-optimized ML computational photography and computer vision algorithms. We also collaborate with the Family of Apps (Facebook, Instagram, WhatsApp, etc.) for enhancing photo and video quality at scale.
Required Skills:
AI Specialist - Computational Photography Responsibilities:
Develop novel ML-based computational photography and image processing algorithms for AR and VR devices, particularly for ML-based denoising applications.
Perform task-specific model optimization (particularly on model architecture) for on-device compute platforms
Drive end-to-end (research to production) ownership of algorithm development
Minimum Qualifications:
Minimum Qualifications:
5+ years of industry experience with real-time (streaming video) ML-based computational photography algorithm development.
5+ years of industry experience with deploying ML computational photography algorithms on smartphones, AR/VR, or other compute and power constrained mobile devices.
5+ years of industry experience with ML image processing algorithms that tap in and out of on-device image signal processors (e.g. Qualcomm or MediaTek ISP)
4+ years of industry experience with CMOS sensor design and deep understanding of CMOS sensor mechanism and architecture.
Strong understanding of mobile AI accelerators (e.g. Qualcomm HTP, DSP, GPU) and can articulate optimal neural network architecture design choices for such accelerators
Strong understanding of engineering design tradeoffs for balancing image processing workloads on various SoC-based processors (e.g. what kind of workloads are best suited for CPU vs GPU vs DSP)
Preferred Qualifications:
Preferred Qualifications:
Shipped ML computational photography features that have deployed commercially on camera-embedded devices, in particular ML-based denoising features.
Published papers in computer vision or computational photography at top-tier conferences (CVPR, ECCV, ICCV, NeurIPS, etc.)
Proficient in low-level device-side production languages such as C/C++, Halide, OpenGL/OpenCL, and Hexagon DSP programming.
Industry: Internet