Applied Research Scientist

AMD AMD · Semiconductors · San Jose, CA · Engineering

Research Scientist at AMD focused on training and advancing large language models (LLMs), large multimodal models (LMMs), and image/video generation models. The role involves exploring novel architectures, large-scale training techniques, pre-training, fine-tuning, and alignment, with a goal to improve state-of-the-art performance and accelerate training/inference speeds. The position also aims to influence AMD's AI platform direction and publish research.

What you'd actually do

  1. Train and finetune LLMs, LMMs, and image/video generation models.
  2. Improve on the state-of-the-art LLMs, LMMs, and image/video generation models.
  3. Accelerate the training and inference speed of LLMs, LMMs, and image/video generation models.
  4. Research novel ML techniques and model architectures.
  5. Influence the direction of AMD AI platform.

Skills

Required

  • PhD or master’s degree or equivalent in machine learning, computer science, artificial intelligence, or a related field.
  • Expertise and hands-on experience on training LLMs, LMMs, and/or diffusion models.
  • Familiar with hyper-parameter tuning techniques, data preprocessing, tokenization methods and latest training approaches for LLMs, LMMs, and diffusion models.
  • Knowledgeable with latest transformer architectures.

Nice to have

  • Experience in developing and debugging in Python.
  • Experience in ML frameworks such as PyTorch, JAX or TensorFlow.
  • Experience with distributed training.
  • Expertise on LLM/LMM/Diffusion pretraining, finetuning, and/or RLHF.
  • Familiar with transformer architecture.
  • Strong communication and problem-solving skills.
  • Publication at top-tier venues is a huge plus.

What the JD emphasized

  • training large language models
  • large multimodal models
  • image/video generation models
  • advance the state-of-the-arts
  • pre-training
  • fine-tuning
  • aligning large language models
  • latest progress and trends
  • novel LLM/LMM and image/video generation architectures
  • large-scale training techniques
  • latest transformer architectures

Other signals

  • training large language models
  • large multimodal models
  • image/video generation models
  • advance the state-of-the-arts
  • pre-training
  • fine-tuning
  • aligning large language models
  • latest progress and trends