Ahmed Haj Ahmed

Intelligence is an emergent property of systems that learn. I'm interested in building those systems.

My work sits at the intersection of research and engineering—developing evaluation frameworks, training novel models, and building infrastructure that pushes the boundaries of what AI systems can understand and generate.

Research Interests

Reasoning & Model Evaluation

Developing frameworks to measure and improve how language models reason, decompose problems, and chain thoughts. Building synthetic data to teach models structured thinking.

Multilingual Evaluation & Cross-Lingual Transfer

Developing evaluation frameworks and transfer learning techniques that enable models to generalize across languages and cultural contexts.

Multimodal AI

Audio-based generative AI, cross-modal understanding, and developing datasets for under-explored modalities in state-of-the-art models.

Experience

Software Engineering Intern

Google · Summer 2025

Building synthetic reasoning chains to teach models how to think.

Software Engineering Intern

ASAPP · Fall 2024

Teaching language models to extract structure from chaos.

Machine Learning Research Assistant

University of Maryland · Summer 2024

Probing what audio models really understand beyond transcription.

Machine Learning Research Assistant

Brown University & Google Research · Spring 2024

Learning representations that transfer across linguistic boundaries.

Machine Learning Research Assistant

Haverford College · Summer 2023

Teaching models to segment images without supervision or labels.

Publications

CuLEmo: Cultural Lenses on Emotion

ACL 2025 (Main Conference)

Benchmarking LLMs for Cross-Cultural Emotion Understanding. T. D. Belay, Ahmed Haj Ahmed, A. Grissom II, et al.

Navigating Dialectal Bias and Ethical Complexities in Levantine Arabic Hate Speech Detection

WACL-4 @ COLING 2024 (First Author)

Ahmed Haj Ahmed, R. Yew, X. Minocher, S. Venkatasubramanian.