Avani Gupta

Avani is an incoming PhD student at Witt Lab and BOLD, supervised by Christian Schroeder de Witt and Jakob Foerster. She is currently interested in alignment and value training, including pro-sociality in multi-agent systems. Previously she researched peer learning in multi-agent systems, concept-based interpretability and control, and worked on training LLMs and building agentic safety pipelines in industry.