"Invent and publish advanced diffusion, transformer, and multimodal text-to-image/video models, including high-resolution generation (up to 16K). Build innovative features for image/video retouching, effects, quality enhancement, and avatar generation. Develop and optimize training and inference pipelines with use of data parallelism, quantization, distillation, TensorRT. Design, build, and maintain training and evaluation workflows using PyTorch 2.x, DataBricks, and SLURM. Create human-in-the-loop evaluations including side-by-side visual comparisons and technical metrics calculation. Collaborate closely with product, design, and engineering teams to ship research-driven features at scale."
"Ph.D. (or equivalent research experience) in Computer Science, Electrical Engineering, Mathematics, or a related field. 5+ years of experience delivering computer vision and generative AI models in production. First-author publications or patents in top-tier venues such as CVPR, ICCV, ECCV, SIGGRAPH, NeurIPS, or ICLR. Deep expertise in PyTorch and CUDA. Hands-on experience with diffusion models, GANs, and vision-language architectures. Strong foundations in linear algebra, probability, optimization, and"
Invent and publish advanced diffusion, transformer, and multimodal text-to-image and video models, including high-resolution generation up to 16K. Build features for image and video retouching, effects, quality enhancement, and avatar generation. Develop and optimize training and inference pipelines using data parallelism, quantization, distillation, and TensorRT. Design, build, and maintain training and evaluation workflows using PyTorch 2.x, DataBricks, and SLURM. Create human-in-the-loop evaluations with side-by-side visual comparisons and technical metrics. Collaborate with product, design, and engineering teams to ship research-driven features at scale. Tune and train custom LLMs for an AI assistant using models like Qwen, LLaMA, and GLM. Require Ph.D. or equivalent, 5+ years production experience, top-tier publications or patents, deep PyTorch and CUDA expertise, and hands-on experience with diffusion models, GANs, and vision-language architectures.
Read at Berlin Startup Jobs
Unable to calculate read time
Collection
[
|
...
]