Research Engineer at Pyka

Candidates should possess strong passion for technology and a desire to influence the real world through algorithms and tools, demonstrating proactivity, self-driven problem-solving, and excellent engineering practices. Proficiency in PyTorch with end-to-end experience in data processing, model training, and deployment is required. Practical experience with LoRA, fine-tuning, DreamBooth, or similar techniques in video/image domains, along with a strong understanding of special effects application scenarios, is essential. Familiarity with video generation and visual models such as ViT, Diffusion, ControlNet, StyleGAN, and a solid understanding of visual generation model training and deployment pipelines are necessary. Solid coding skills and a strong grasp of algorithm fundamentals are required, with experience in distributed training tools like FSDP or Deepspeed being a plus. Experience in large-scale data processing, dataset construction, and automated data cleaning is a plus, as is experience with AIGC, video effects, or asset generation products.