Potential Topics

These are some topics that interest me. But I am open to new ideas and topics. In fact, a major portion of your research time will be spent on finding good problems to work on.

  1. Representation Learning and Computational Linguistics — text embeddings, interpretability, multilingual representations
  2. Reinforcement Learning for LLM Reasoning — reward design, process supervision, search
  3. Safety and Alignment — scalable oversight, tiered alignment, agent safety
  4. Language Agents — web navigation, tool use, retrieval-augmented generation, hallucination mitigation
  5. Multimodal Understanding — language and perception, vision-language reasoning, world models
  6. New Architectures and Training Objectives — mixture of experts, positional embeddings, pre-training and fine-tuning objectives

Some of these topics can be co-supervised with other faculty at MILA and McGill.

Site Counter