HE PERSON:
This role is for you if you are passionate about reading through the latest literature, coming up with novel ideas, and implementing those through high quality code to push the boundaries on scale and performance. The ideal candidate will have both theoretical expertise and hands-on experience with developing LLMs, LMMs, and/or diffusion models. We are looking for someone who is familiar with hyper-parameter tuning methods, data preprocessing & encoding techniques and distributed training approaches for large models.
KEY RESPONSIBILITIES:
- Pre-train and finetune over large GPU clusters while optimizing for various trade-offs.
- Improve upon the state-of-the-art in Generative AI model architectures and training techniques.
- Accelerate the training and inference speed across AMD accelerators.
- Publish your work at top-tier conferences & workshops and/or through technical blogs.
- Engage with academia and open-source ML communities.
- Drive continuous improvement of infrastructure and development ecosystem.
PREFERRED EXPERIENCE:
- Strong development and debugging skills in Python.
- Experience in deep learning frameworks (like PyTorch or TensorFlow) and distributed training tools (like DeepSpeed or Pytorch Distributed).
- Experience with fine-tuning methods (like RLHF & DPO) as well as parameter efficient techniques (like LoRA & DoRA).
- Solid understanding of various types of transformers and state space models.
- Strong publication record in top-tier conferences, workshops or journals.
- Solid communication and problem-solving skills.
ACADEMIC CREDENTIALS:
- Advanced degree (Master's or PhD) in machine learning, computer science, artificial intelligence, or a related field is expected. Exceptional Bachelor's degree candidates will also be considered.
