Teams at Krea

Research and engineer large-scale image and video diffusion models. Train and optimize distributed models on GPU clusters, implement parallelism strategies, profile and debug training infrastructure, improve data and model quality, design experiments and evaluations, and develop fault-tolerant solutions for hardware and numerical issues. The role requires rapid iteration, strong research judgment, custom data pipeline development, and hands-on expertise with PyTorch, distributed training, low-precision computation, and diffusion training across pretraining, preference optimization, and reinforcement learning.

Research and develop posttraining methods for large-scale image and video diffusion models. Responsibilities include fine-tuning, reinforcement learning, preference optimization, distillation, reward modeling with VLMs, LLM prompt expansion, evaluation design, safety alignment, distributed training, and inference optimization. The role partners with data, research, and engineering teams to improve model quality, aesthetics, controllability, and product integration.

Build and improve AI creative products using SvelteKit or React, including real-time collaboration, canvas tooling, durable workflows, public APIs, and observability. The role may also involve Python ML inference on GPU clusters and multi-cloud Kubernetes infrastructure. Engineers own features end-to-end, from ideation through launch and measuring impact, while contributing to robust, revenue-driving products for creative users.