Soujanya Poria
সৌজন্য পড়িয়া
Associate Professor, Nanyang Technological University (NTU)
School of Electrical and Electronic Engineering
College of Computing and Data Science (CCDS) (courtesy appointment)
I lead the DeCLaRe Lab (Deep Cognition and Language Research). My current research focuses on memory for AI agents, continual learning, reliable multi-agent systems, and efficiency.
These questions run through the lab’s work on AI safety and trustworthiness, multimodality, AI for science, and embodied AI.
Email Lab Website Google Scholar DBLP ACL Anthology GitHub Twitter
Research
Memory for AI agents
What an agent should retain, how that memory updates, and how it shapes the next action.
Continual learning
How a deployed model keeps learning without losing what it already had.
Reliable multi-agent systems
How agents learn whom to trust and when to ask for advice.
Efficiency
How much computation a model needs to learn and to retain what is useful.
News
I’m excited to be one of the PC Chairs for EMNLP 2027.
I will also co-chair the Demonstration Track at ACL 2027.
Beyond Better Actions: What Generalist Physical Agents Still Need
Invited talk at the IMDA Technical Sharing Session and AI Wednesdays, Singapore. October 2026.
Embodied AI for Human-Robot Collaborative Task Execution
A new project with CNRS@CREATE and the National Research Foundation, 2026–29.
2 papers accepted to NeurIPS 2026, including 1 Spotlight
2 papers accepted to ICML 2026
4 papers accepted to ICLR 2026
Highly Cited Researcher
Recognized by Web of Science, 2026.
VentureBeat features δ-mem
An interview on working memory for long-running AI agents. May 2026.