Five researchers affiliated with the Center for Language and Speech Processing have been selected for 2025–2026 Faculty Research Awards from the Johns Hopkins University + Amazon Initiative for Artificial Intelligence (AI2AI). Their projects span multimodal speech synthesis, inference-time scaling, retrieval-augmented generation, 3D vision-language reasoning, and structured data understanding.

The AI2AI Faculty Research Awards support innovative projects across Johns Hopkins that push the boundaries of artificial intelligence in real-world applications. Below are the CLSP-affiliated awardees for this year’s cycle.

---

Daniel Khashabi, Assistant Professor, Department of Computer Science

Project: Multi-modal Inference-time Scaling
Khashabi’s work aims to unify language-based and vision-based inference-time scaling, enabling more powerful and efficient multimodal reasoning.

---

Philipp Koehn, Professor, Department of Computer Science

Project: Reasoning over Information in Tabular Data
Koehn is developing new benchmarks and methods that improve how AI systems understand and reason over structured tabular data.

---

Berrak Sisman, Assistant Professor, Department of Electrical and Computer Engineering

Project: Multimodal Speech Synthesis: Leveraging Computer Vision and LLMs for Expressive Voice Generation
Sisman’s project explores multimodal approaches that integrate text, images, and gestures to produce more natural and expressive synthesized speech.

---

Mahsa Yarmohammadi, Assistant Research Scientist, Center for Language and Speech Processing

Project: Enhancing Retrieval-Augmented Generation with Multi-Vector Representation
Yarmohammadi is improving retrieval-augmented generation systems by introducing multi-vector representations that preserve richer semantic context.

---

Alan Yuille, Bloomberg Distinguished Professor of Cognitive Science and Computer Science

Project: Next-Generation 3D Vision-Language Models with Physics, Counterfactual Reasoning & Planning
Yuille’s research develops next-generation models capable of understanding dynamic 3D/4D scenes using geometry, physics-based reasoning, and counterfactual exploration.

---

A full list of 2025–2026 AI2AI award recipients and project summaries is available on the program’s webpage.