CV

Ph.D.

Columbia University in the City of New York

Computer Science, Natural Language Processing

• Advisor: Prof. Julia Hirschberg

M.Sc.

Columbia University in the City of New York

Machine Learning

B.Math

University of Waterloo

Computer Science, Statistics, Actuarial Science

• Graduated with Distinction

Senior Applied Scientist

Microsoft, IDEAS Research

Led by Dr. Scott Counts

• Advancing research on productivity agents that use workflow, workstream, artifact, collaborator, and context signals beyond isolated task-level assistance.
• Developed telemetry-driven workflow understanding methods that convert Microsoft 365 user actions into hierarchical natural-language representations, enabling analysis across tasks, sessions, work episodes, and long-running workstreams.
• Delivered research prototypes, internal demos, technical writeups, and evaluation frameworks for context-aware agent experiences, including workflow-centered memory, workstream discovery, and telemetry-based diagnosis of Copilot-assisted workflows.

Research Scientist Intern

Adobe Research, Data Intelligence Team

Led by Dr. Vishy Swaminathan

• Developed SteER, an interactive deep research agent that adaptively pauses to solicit user guidance, enabling more personalized and controllable long-form research workflows.
• Designed a cost-benefit pause policy and persona-aware branch selection method to balance user alignment, information gain, exploration diversity, and interaction cost.
• Delivered research prototypes, internal demos, human evaluation studies, and invention disclosure materials demonstrating improved alignment and focus over autonomous deep research baselines.

AI/ML Data Associate Research Intern

JP Morgan Chase, Machine Learning Center of Excellence

Led by Dr. Lidia Mangu

• Developed NovAScore, an automated document-level novelty metric that combines atomic content units with salience weighting to evaluate information distinctiveness and reduce redundancy.
• Delivered a patent-pending framework for document curation, training data selection, and retrieval/evidence ranking.

Research Scientist Intern

Meta, Multimodal Team

Led by Dr. Florian Metze

• Built multimodal sentiment classification models for short-form videos, improving cross-modal representation learning and achieving competitive performance with state-of-the-art baselines.
• Designed a multimodal BYOL self-supervised learning framework to improve cross-domain representation adaptation.

Ph.D. Researcher

Columbia University

Advisor: Prof. Julia Hirschberg; Dissertation: Towards Trustworthy AI: Detecting, Understanding, and Mitigating Information Disorder

• Developed a detection-understanding-mitigation framework for trustworthy AI systems addressing information disorder across social media, multimodal content, and LLM-mediated communication.
• Built models and datasets for misinformation, propaganda, radicalization, malicious intent, and audience perception, integrating textual, visual, social-network, and human-centered signals.
• Designed mitigation systems for LLM-era risks, including social-engineering defense, factuality-controlled generation, distinctiveness-aware curation, and human-in-the-loop agent steering.