Research Scientist – Efficient and Controllable Generative Models

Huawei's Computer Vision and Machine Learning Lab in Zurich is expanding its research team with a Research Scientist role. Huawei, a global leader in ICT, builds AI‑driven cloud services, smart devices and telecom infrastructure that reach more than one‑third of the world's population. The lab focuses on generative modeling for images, video and 3D/4D content, aiming to embed these capabilities into millions of consumer devices such as smartphones, tablets, laptops and autonomous‑vehicle platforms. In this position you will spend each day designing and testing new generative approaches that meet strict on‑device constraints. Your work will cover diffusion and flow‑based models, few‑step or one‑step generation, model distillation, quantisation and hardware‑aware architecture design. You will create techniques that keep generated output faithful to the input and user intent, preventing artifacts like distorted faces or flickering frames. The research will be extended to 3D and 4D scenarios, including neural and Gaussian scene representations, and will be integrated into camera pipelines, image/video enhancement, editing tools and real‑time rendering workflows. Collaboration with product, hardware and software engineers will turn prototypes into shipped features, while you also publish results at top conferences and mentor junior team members. The role requires a PhD, or equivalent research experience, in computer vision, graphics, machine learning or a closely related discipline. A strong record of publications in generative modeling, image/video restoration, neural rendering, 3D/4D vision or efficient deep learning is essential. You must be comfortable writing clean, high‑performance research code using modern deep‑learning frameworks such as PyTorch, and possess a deep technical understanding of diffusion and flow‑based models, including sampling, guidance, conditioning and distillation. The description also highlights the need for curiosity, independence and a drive to translate research into real‑world impact. Preferred experience includes work on controllable and faithful generation (e.g., preserving identity or text), deployment of models on mobile or embedded platforms (NPUs, quantisation‑aware training, on‑device runtimes), familiarity with 3D data structures like meshes, point clouds, implicit fields or Gaussian splatting, and a background in autonomous‑driving perception or simulation. The team is international, based in the heart of Zurich, a city renowned for its vibrant research ecosystem and high quality of life, offering a collaborative environment that bridges academic excellence and product development. What the role asks for: - PhD in computer vision, graphics, machine learning or related field - Strong publication record in generative modeling or related areas - Hands‑on experience with modern deep‑learning frameworks (e.g., PyTorch) - Deep understanding of diffusion and flow‑based models - Demonstrated curiosity, independence and impact‑driven mindset - Experience with controllability and faithfulness in generation (nice to have) - Experience deploying models on mobile or embedded platforms (nice to have) - Familiarity with 3D representations such as meshes or Gaussian splatting (nice to have)