P Sam Sahil

P Sam Sahil

Research Intern • Research Assistant

Research Intern at Northwestern University, NTU Singapore, and University of Hamburg.

Pivoting from black-box models to principled design. Focused on building systems that are not just performant, but reasoning-capable and transparent.

Research Focus

Interpretability & Trust

Moving beyond "black box" optimization to design transparent AI architectures. Investigating how internal reasoning mechanisms can be made explicit and human-verifiable.

Neuro-Symbolic AI

Bridging the gap between the scale of neural networks and the logic of symbolic reasoning. Developing dual-path architectures (GAT + RoBERTa) to ground semantics in structured knowledge.

Multilingual NLP

Breaking language barriers in emotion detection and polarization. Focusing on cross-lingual generalization to capture nuance across diverse cultural contexts and low-resource languages.

Comp. Social Science

Applying rigorous NLP pipelines to large-scale social data. From geo-tagging historical newspapers to analyzing online polarization to answer fundamental questions.

Latest Updates

Present Event

POLAR @ SemEval-2026

Co-organizer for Task 9: Detecting Multilingual, Multicultural, and Multievent Online Polarization.

Visit polar-semeval.github.io →
Jan 2026 Internship

Northwestern University

Incoming Research Assistant Intern at Change Lab (Kellogg School of Management).

Jan 2026 Internship

NTU Singapore

Incoming Research Intern working on Fast RAG on Edge Devices.

Selected Publications

PUBLISHED (ACL)

Breaking Language Barriers in Emotion Detection

SemEval-2025 Task 11 • Team A

Developed multilingual models for multi-label emotion detection across diverse languages, achieving competitive leaderboard ranks (Russian 7th, Hindi 9th).

Multilingual NLP Emotion AI
UNDER REVIEW

Synergizing Contextual Semantics and Moral Knowledge Graphs

Dual-Path Architecture for Moral Foundation Prediction

Proposed a hybrid RoBERTa + GAT architecture leveraging eMFD knowledge graphs with cross-attentive fusion to improve moral foundation prediction.

Graph Neural Networks Moral AI