Harm de Vries

Harm de Vries

Staff Research Scientist

ServiceNow, Amsterdam

About me

I’m a Staff Research Scientist at ServiceNow and lead of its Large Language Models Lab. My lab works on the responsible development of generative AI models, with a focus on natural language and source code. We take the lead in open-scientific collaborations, such as BigCode, wherein we develop large language models from scratch and establish the foundations of ethical data governance practices at scale.

I obtained a PhD from Mila (Université de Montréal) under the supervision of Aaron Courville. In my thesis, I investigated deep learning methods for visual dialogue applications. I was also fortunate enough to spend time at the research labs of Facebook, Twitter, and INRIA.

  • Efficient training and inference of LLMs
  • Adaptation methods for LLMs
  • Ethical data governance at scale
  • Interactive code generation
  • Conversational question answering
  • Task-oriented dialogue
  • PhD in Artificial Intelligence, 2019