I'm the cofounder of Krea (https://www.krea.ai/), a startup in San Francisco building browser-based creative tools and AI systems focused on personalization, controllability, aesthetics, and speed. We are also working real-time video models and world models.
Some customers: Pixar, Shopify, Fox News, Amazon Studios.
We are ~15 technical staff members (https://www.krea.ai/careers) looking for others interested in tackling challenges in the creative tooling (with AI) space. We have a lot of GPUs and plan on getting more.
Some challenges: large-scale training of diffusion models; post-training such models using techniques like DPO; UI testing framework to simulate user actions; distillation of state-of-the-art models like Flux to achieve real-time performance at inference-time; zero-downtime and blue-green deployments; recommendation systems for user-content; vector-based search API for filtering through millions of user assets; multi-cloud multi-cluster infrastructure management; enterprise-grade granular access control similar to how Linux ACLs work.
Some investors: cofounder of Meta/Facebook AI Research, founding member of OpenAI, a16z.
Our tech stack: CUDA, PyTorch, Kubernetes, Tigris, Node, Svelte(Kit), WebGL/WebGPU/WASM, Python, Ceph.
If you want to learn more or meet up to talk about creative tooling or research, email me at d@krea.ai.