Member of Technical Staff - Image / Video Generation
Black Forest Labs is seeking a Member of Technical Staff specializing in image and video generation to join our pioneering team. As the creators of Latent Diffusion, Stable Diffusion, and FLUX, we are at the forefront of generative AI technologies that empower millions of creators and businesses worldwide. Headquartered in Freiburg, Germany, with a growing presence in San Francisco, we are committed to research excellence and open science to expand human creativity.
In this role, you will train large-scale diffusion transformer models for image and video data, focusing on empirical experimentation to identify impactful architectural choices. Your responsibilities include conducting rigorous ablation studies to isolate variables and control for confounds, thereby producing trustworthy insights that guide our research direction. Additionally, you will evaluate the speed-quality trade-offs of neural network architectures in production settings and fine-tune diffusion models for specialized applications such as upscalers and inpainting/outpainting models.
The ideal candidate has hands-on experience training large-scale diffusion models for image and video data, with a deep understanding of common failure modes and critical training factors. Proficiency in PyTorch, transformer architectures, and modern deep learning ecosystems is essential. A solid grasp of distributed training techniques, including Fully Sharded Data Parallel (FSDP), low precision training, and model parallelism, is required. Experience in fine-tuning diffusion models for specialized applications and effectively evaluating image and video generative models is highly valued.
We offer a collaborative and innovative work environment grounded in our core values: obsession with research excellence, low ego, boldness in taking ambitious bets, and kindness in interpersonal interactions. Our distributed team operates with flexibility, encouraging in-person collaboration in our Freiburg or San Francisco offices at least two days a week, or one full week every other week, with reasonable travel costs covered. This structure ensures meaningful in-person engagement while accommodating diverse working preferences.