Role Overview: Design and build CNN/UNet-based generative models, working with GAN and diffusion architectures at extremely high resolutions. You will collaborate with a team to make architecture decisions, implement custom layers and loss functions, and deliver proof-of-concept demos demonstrating new network designs. Responsibilities: Architect end-to-end CNN and UNet models for image generation. Design and implement custom convolutional blocks and loss functions. Develop and maintain GAN- and diffusion-based pipelines capable of producing extremely high-resolution outputs. Translate model diagrams into layer-level code; visualize feature maps and activations to validate design choices. Prototype and iterate on new model ideas, producing working demos of each design. Collaborate with research and engineering peers to refine and scale generative pipelines. Must-Have Qualifications: Deep learning experience with custom CNN and Unet architectures. Proven ability to build models layer-wise: modify layers, inspect internals, and adjust architectures. Hands-on experience designing and training GANs and diffusion models. Experience writing and integrating custom loss functions for image-generation tasks. Please Read Qualifications: Portfolio with at least two projects or repositories demonstrating: Custom CNN/Unet architectures designed or modified from scratch. Layer-level modifications and custom losses implemented. GAN or diffusion-based work that produced high-resolution outputs. Application Requirements: Links to two past projects (GitHub, GitLab, or similar) showing your custom CNN/Unet/GAN/Diffusion work, with brief notes on your contributions. Will be required to demo and create a small POC as proof of skills. We look forward to meeting eager individuals. Show more Show less