Research Scientist, ML Recommendation Systems, Applied Machine Learning Team

ByteDance ByteDance · Big Tech · San Jose, CA · R&D

Research Scientist role focused on building and scaling machine learning models for recommendation systems, including end-to-end generative systems and reinforcement learning for personalization. The role involves researching and applying multi-modal techniques, optimizing model architectures for large-scale training and inference, and collaborating with product and engineering teams for deployment. Requires expertise in deep learning, LLMs, multi-modal learning, and production ML pipelines, with a strong publication record.

What you'd actually do

  1. Build and scale up machine learning models for recommendation systems
  2. Research and apply multi-modal techniques (leveraging text, image, video) to create a holistic understanding of content and user preferences
  3. Pioneer new modeling strategies by researching and integrating long-term user behavior signals to drive sustained engagement and satisfaction, by using techniques such as reinforcement learning
  4. Partner closely with the infrastructure team to co-design and optimize next-generation recommendation model architectures and systems, ensuring high-performance, low-latency, and cost-efficient training and inference at a massive scale.
  5. Work hand-in-hand with product, engineering, and design teams to rigorously test and deploy end-to-end solutions, validating their impact and ensuring they create a seamless and enhanced user experience.

Skills

Required

  • Python
  • C++
  • PyTorch
  • TensorFlow
  • Transformers
  • Large Language Models (LLMs)
  • multi-modal learning
  • end-to-end ML pipelines
  • production environment

Nice to have

  • Ph.D. in a relevant field
  • recommendation systems
  • reinforcement learning

What the JD emphasized

  • scaling machine learning models
  • multi-modal techniques
  • reinforcement learning
  • massive scale
  • end-to-end solutions
  • track record of publications at accredited peer-reviewed conferences such as NeurIPS, ICML, ICLR, KDD, RecSys, WWW

Other signals

  • recommendation systems
  • large scale
  • production deployment
  • multi-modal
  • reinforcement learning