Diffusion AI Researcher — Image & Video at Scale

AI Chopping Block; San Francisco, Northern (CA, KY); Hybrid; USD 150,000 - 210,000; Full time

Benefits offered by this job

Health insurance

401k with 4% company match

Meals in the office

Transit covered

Visa sponsorship for international位

Job summary

Krea is building next-generation AI creative tools and seeks an experienced Researcher with engineering skills to advance large-scale image and video model training experiments. You will train diffusion models, optimize distributed runs, and push model quality while profiling architectures, loaders, and memory systems.

The role emphasizes hands-on experimentation, fast iteration, and collaboration with a talented team in a research-driven environment.

Qualifications

• Proven track record in working with image or video models at scale.
• Strong proficiency in PyTorch.
• Experience with distributed training paradigms such as FSDP, CP, SP, TP, and EP.
• Ability to profile and debug large distributed training jobs.
• Good knowledge of low precision training (FP8, NVFP4, MXFP8).
• Solid understanding of diffusion model training pipelines across pretraining to reinforcement learning.
• Keeping up with developments in LLM, VLM, representation learning, and robotics research.

Responsibilities

• Train diffusion models for image and video generation on large GPU clusters.
• Fully optimize and profile large distributed training runs across model architectures, kernels, data loading, memory, and communication.
• Implement and improve various distributed training strategies including FSDP, CP, SP, TP, and EP.
• Continuously improve model quality and reliability through data, architecture, and training pipeline design.
• Debug distributed training errors and implement fault tolerance solutions, identifying hardware and software bottlenecks.
• Ablate architecture, attention, optimizer, data, and algorithm choices to improve efficiency and performance.

Skills

PyTorch

Distributed training

Image/Video models

Profiling & debugging

Low precision training

Diffusion models

Tools

GPU clusters

NVLink

Infiniband

NCCL

Job description

Krea is building next-generation AI creative tools and seeks an experienced Researcher with engineering skills to advance large-scale image and video model training experiments. You will train diffusion models, optimize distributed runs, and push model quality while profiling architectures, loaders, and memory systems.

The role emphasizes hands-on experimentation, fast iteration, and collaboration with a talented team in a research-driven environment.