Experience
Alexandria
Team, Microsoft Research, London, UK
MSR x Mila Research Collaborator • Feb. 2023 to preset
Propose a multi-stage architecture for extracting structured kowledge from unstructured
texts using LLMs.
Propose a unified training paradigm with diffusion model applicable to various
information
retrieval tasks (i.e., KB Completion, Entity Linking, Information Extraction,
KB-augmented Q&A).
Advisor: Bhaskar
Mitra
|
|
|
Descartes Team,
Google Research, Mountain View, USA
Student Researcher • Aug. 2023 to Feb. 2024
Investigate knowledge localization in LLMs for both dense architectures and sparse
architures (e.g., Mixture-of-Experts).
Advisor: Cicero
Nogueira dos Santos
|
|
|
MuDcATs (Making Decisions with Activity-based, Temporal & Sequential Data) Team,
Google Research, Mountain View, USA
Research Intern • Nov. 2022 to Feb. 2023
Propose density-based user representation (DUR) through Gaussian Process Regression for
better personalized multi-interest retrieval.
Advisor: Craig Boutilier
|
|
|
FATE
(Fairness, Accountability, Transparency, and Ethics) Team., Microsoft
Research, Montreal, Canada
MSR x Mila Research Collaborator • Sep. 2020 to Jun. 2021
Propose a multi-objective optimization framework for fairness-aware recommendation,
Multi-FR, that adaptively balances accuracy and fairness for various stakeholders with
Pareto ptimality guarantee.
Advisor: Fernando Diaz
|
|
|
|