Skip to content
Khalid Saifullah
GitHubXEmail

Khalid Saifullah

Ph.D. student at the University of Maryland

I study language-model systems as a way to understand what the real building blocks of general intelligence might be. My work asks how models can do more with less compute and data, how we can tell when their apparent reasoning is reliable, and how to keep powerful systems useful, steerable, and safe under pressure.

I am motivated by the possibility that efficient and trustworthy AI can expand human agency rather than replace it: systems that retrieve grounded evidence, evaluate themselves honestly, resist brittle failure modes, and help people make progress on problems that matter. I am advised by Tom Goldstein at UMD, spending Summer 2026 at Microsoft AI, and previously interned at Adobe Research.

EmailGoogle ScholarLinkedInTwitterResearch

Selected Publications

ICLR SCI-FM 2025
Retrieval models

Scalable Pretraining of Retrieval Models

Khalid Saifullah, John Kirchenbauer, Rifaa Qadri, Abhimanyu Hans, Neel Jain, Yuxin Wen, Siddharth Singh, Brian R. Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein

Full list on Google Scholar.

Experience

Research Intern, Microsoft AI

Summer 2026

Working on frontier AI systems during the summer internship.

Research Scientist Intern, Adobe Research

2025 and 2023

Worked on machine learning research across evaluation, robustness, and generative systems.

Selected Talks