About the job
The Vision-Applied Research team focuses on applied research in Generative AI and CV/Multimodal Understanding, and delivering intelligent solutions to ByteDance products, e.g., TikTok, CapCut, and Lemon8, enabling users to make and share creative content in a much easier way. The team has research groups dedicated to generative models for content creation, image generation, video synthesis, intelligent image/video editing, and virtual humans. We are seeking an experienced Multimodal Model Training and Inference Optimization Engineer with expertise in optimizing AI model training and inference, including distributed training/inference and acceleration. The ideal candidate will work at the cutting edge of AI efficiency, enhancing the performance, scalability, and deployment of large-scale generative AI models.
Responsibilities
- Optimize large model training pipelines to improve efficiency, speed, and scalability.
- Develop and improve distributed training strategies such as data parallelism, model parallelism, pipeline parallelism and communication to accelerate model training.
- Benchmark and profile deep learning models to identify performance bottlenecks and optimize computational resources.
Qualifications
Minimum
- M.S or PhD in Computer Science, Electrical Engineering, Artificial Intelligence, or a related field.
- Experience in AI model training optimization.
- Strong software engineering skills, including proficiency in Python, C++, and CUDA.
- Strong proficiency in deep learning frameworks such as PyTorch, Megatron and Deepspeed.
- Experience with distributed training techniques such as data parallelism, model parallelism, and pipeline parallelism.
- Knowledge of transformers and diffusion models.
Preferred
- Candidates with publications at conferences such as MLSys, NeurIPS, ICLR, or ICML are preferred.
- Strong communication and teamwork skills.
- Self-motivated and strong problem-solving skills.
- Ability to work collaboratively in multi-functional teams.
- Experienced in implementing and optimizing complex and performance-critical systems.