CubiCasa

LLM Engineer

• Inference pipelines: Implement modular inference pipelines using SOTA Latent Diffusion Models to facilitate complex image-to-image pipelines.
• Advanced Inpainting & Editing: Develop robust algorithmic solutions for targeted inpainting, outpainting, and coherent background generation within existing scenes.
• Semantic Processing: Integrate SOTA segmentation models to automatically identify and isolate specific image regions for targeted visual modification.
• Complex Chaining: Utilize custom Python backends to chain non-linear workflows, combining detection, segmentation, and generation steps.
• Geometric & Structural Conditioning: Implement and fine-tune structural conditioning preprocessors to ensure generated elements strictly respect the vanishing points, perspective, and geometry of the source input.
• Style Adaptation: Train and maintain parameter-efficient fine-tuning (PEFT) adapters to steer models toward specific high-value aesthetic styles and lighting conditions without catastrophic forgetting.
• Automated Quality Scoring: Develop AI-on-AI scoring systems using advanced computer vision metrics to detect hallucinations, artifacts, or perspective failures before human review.

Generative AI Engineer