Announcing the 2024 Boston University BU-AIRR Award Recipients
By Maureen Stanton
The Hariri Institute for Computing is pleased to announce the recipients of the 2024 BU AI Research Resource Program (BU-AIRR) awards. This program aims to enable and seed large-scale AI research at Boston University by funding access to the powerful production cloud resources of the New England Research Cloud (NERC).
“AI research is resource intensive, often requiring enormous computational resources to analyze vast datasets, train large deep neural network models, and perform complex calculations,” says Yannis Paschalidis, director of the Hariri Institute for Computing. “While cloud computing offers opportunities to use powerful computing capabilities without the need to maintain a hardware infrastructure, most commercially-offered GPU as a Service (GPUaaS) solutions are very costly and beyond the reach of individual research groups.”
The NERC is an affordable fee/hour production cloud service, operated by BU and Harvard Research IT, that offers access to top-of-the-line CPUs (Nvidia A100 GPUs, and recently acquired Nvidia H100 GPUs) and storage solutions. Administered by the Mass Open Cloud (MOC) Alliance, the NERC was founded to provide more cost-effective high-end cloud services to any BU researcher, and researchers from several other institutions.
“The MOC Alliance is excited to offer this unique opportunity to introduce the NERC production cloud to BU researchers, delivering high availability, scalability and reliability to accelerate research goals,” said Nancy Clinton, managing director, MOC Alliance. “The Alliance will also facilitate helping faculty get projects up and running quickly to optimize the performance of their cloud workloads. We look forward to working with the new awardees and exploring new opportunities.”
The 2024 BU-AIRR program awarded $290,000 to 26 AI research projects from 35 BU faculty members affiliated with six colleges and 13 departments. Funding was made possible through contributions from the Hariri Institute, the College of Engineering, the Department of Electrical & Computer Engineering, and the Office of Research. A total of 33 proposals were received and were evaluated by an ad-hoc committee of three senior faculty leaders against selection criteria that included proposal originality, potential for enabling future sponsored research, and need for high-end GPUs.
Learn more about the funded projects below.
1) Project: “A Biomechanically and Anatomically Aware Multi-Modal Foundational Model for the Lung CT Images” focuses on developing an anatomically and biomechanically informed generative foundational model for lung CT images. Personnel: PI: Kayhan Batmanghelich, PhD, Assistant Professor, ENG (ECE).
2) Project: “Mammo-FM: Foundational Vision-Language Model for Enhancing Breast Cancer Risk Prediction Explainability and Bias Mitigation through Clinician-AI Collaboration” focuses on developing “Mammo-FM,” a multi-modal foundational model trained on extensive mammogram-report pairs, capable of addressing existing challenges by building a scalable and equitable solution to enhance breast cancer detection and risk prediction. Personnel: PI: Kayhan Batmanghelich, PhD, Assistant Professor, ENG (ECE).
3) Project: “Machine Learning Assisted Transition Metal Dichalcogenide Dynamical Studies” aims to explore MXenes, an even newer avenue for materials research with exciting applications such as wearable biosensors, batteries, plasmonics, and more. Personnel: PI: David Coker, PhD, Professor, CAS (Chem), ENG (MSE)Collaborating Investigators: Xi Ling, Associate Professor, CAS (Chem) Student: Camilo Zuluaga, CAS (Chem).
4) Project: “AI for Automating Cloud Operations” aims to address key challenges in cloud operations by developing novel solutions for trace analysis, vulnerability detection, and anomaly management. Personnel: PI: Ayse Coskun, PhD, Associate Professor, ENG (ECE, SE); Collaborating Investigators: Gianluca Stringhini, PhD, Associate Professor, ENG (ECE); Manuel Egele, Associate Professor, ENG (ECE), Brian Kulis, Associate Professor, ENG (ECE, CS, SE), CDS.
5) Project: “AI for Improving Efficiency and Sustainability of Computing” aims to advance the energy efficiency and sustainability of HPC systems and data centers by focusing on intelligent resource management, automated performance diagnostics, and adaptive power strategies in HPC systems and data centers. Personnel: PI: Ayse Coskun, PhD, Associate Professor, ENG (ECE, SE); Collaborating Investigators: Manuel Egele, Associate Professor, ENG (ECE), Brian Kulis, Associate Professor, ENG (ECE, CS, SE), CDS; Adam Smith, Professor, CAS (CS, ECE), CDS; Ajay Joshi, Professor, ENG (ECE); Emiliano Dall’Anese, Associate Professor, ENG (ECE, SE).
6) Project: “LLMs for Provably Correct Distributed Programming” aims to design a robust LLM ecosystem to develop and verify safety-critical decision-making systems. Personnel: PI: Ankush Das, PhD, Assistant Professor (CAS-CS); Collaborating Investigators: Shrimai Prabhumoye, PhD, Adjunct Assistant Professor, CAS (CS); Deepti Ghadiyaram, PhD, Assistant Professor, CAS (CS).
7) Project: “Uncovering Dynamic Neural Interactions Between Hierarchical Brain Regions” aims to build a Long Short-Term Memory (LSTM) model, a type of a recurrent neural network, that would infer medulla activity using the keypoint kinematic features recorded via video without the need for direct brain probe measurements. Personnel: PIs: Brian DePasquale, PhD, Assistant Professor, ENG (BME); Michael Economo, PhD, Assistant Professor, ENG (BME).
8) Project: “Understanding Behavioral Dynamics During Learning in Genetically Diverse Populations” aims to utilize both experimental and computational advances to explore the long-term evolution of behaviors during learning across a genetically diverse population. Personnel: PI: Brian DePasquale, PhD, Assistant Professor, ENG (BME); Collaborating Investigators: Jerry Chen, PhD, Associate Professor, CAS (Bio) Student: Halley Jeanne Dante, PhD student, ENG (BME).
9) Project: “Democratizing Research on Multimodal Foundation Models” aims to democratize the research on the pretraining of foundation models so that researchers with university budgets can meaningfully contribute to related open problems, such as the models’ lifelong learning, fairness and safety, and efficiency. Personnel: PI:Boqing Gong, PhD, Assistant Professor, CAS (CS); Collaborating Investigator: Venkatesh Saligrama, PhD, Professor, ENG (ECE, SE).
10) Project: This project aims to develop specialized large language models (LLMs) for the enhancement of binding affinity and classification prediction and to generate accurate structural models for antibody-antigen and MHC-peptide complexes. Personnel: PIs: Diane Joseph-McCarthy, PhD, Professor of the Practice, ENG (BME), CAS Chem); Sandor Vajda, Professor, ENG (BME, SE), CAS (Chem).
11) Project: “How Video Content Drives Sales in Influencer Marketing” seeks to understand the role of short videos, such as those on TikTok, in influencing consumers’ purchase decisions and how businesses collaborate with creators to promote products through such content. Personnel: PI: Shunto Kobayashi, PhD, Assistant Professor, QST (Marketing); Student: Wenyi Huang, PhD student, QST (Marketing).
12) Project: “Leveraging AI and Remote Sensing Data for Automated Flood Mapping” aims to develop an automated method for flood detection and mapping using a combination of optical multispectral and Synthetic Aperture Radar (SAR) images with a deep learning approach. Personnel: PI: Magaly Koch, PhD, Research Associate Professor, CAS (E&E).
13) Project: In this project, a multidisciplinary team is developing an automated tool to address challenges of distinguishing between healthy sharp wave ripple (SWR) from pathologic spike ripples (SR) in human hippocampal recordings. Personnel: PI: Mark Kramer, PhD, Professor, CAS (Math).
14) Project: “Multi-Task Decision Transformers” studies the problem of task alignment and multi-task generalization in Decision Transformers, with applications to offline reinforcement learning. Personnel: PI: Wenchao Li, PhD, Associate Professor, ENG (ECE, SE).
15) Project: “Uncertainty Quantification for Vision-Language Models” aims to develop novel uncertainty quantification methods for vision-language models, with a specific focus on improving the reliability of these models in automatic radiology report generation. Personnel: PI: Wenchao Li, PhD, Associate Professor (ENG-ECE, SE).
16) Project: “LLMKernel: Efficient LLM Pre-Training and Fine-Tuning Strategies for Systems” aims to enhance the scalability and resource efficiency of pre-training and fine-tuning LLMs, particularly in multi-modal settings, through the investigation of effective continual learning approaches that can reduced need to re-train models from scratch and save an order of magnitude of training resources. Personnel: PI: Eshed Ohn-Bar, PhD, Assistant Professor, ENG (ECE); Collaborating Investigators: Lance Galletti, Lecturer, CAS (CS), Kate Saenko, Professor, CAS (CS), Renato Mancuso, Associate Professor, CAS (CS), Adam Smith, Professor, CAS (CS); Students: Lei Lai, PhD student, CAS (CS); Arijit Ray PhD student, CAS (CS); Hee Jae Kim, PhD student, ENG (ECE).
17) Project: “Automated Lesion Segmentation in Traumatic Brain Injury and Relationship with Pupillometry” aims to determine the association between quantitative pupillometry and contusion location/contusion volume, and to better understand which features are associated with actionable and clinically relevant pupillometric thresholds. Personnel: PI: Charlene J. Ong, MD, MPHS, Assistant Professor, Chobanian & Avedisian SOM (Neurology);Collaborating Investigators: Shariq Mohammed, PhD, Assistant Professor, SPH (Biostatistics); Mohamad Abdalkader, MD, Associate Professor, Chobanian & Avedisian SOM (Radiology); and Medical Student: Charles L. Chen, MS, Chobanian & Avedisian School of Medicine.
18) Project: “Learning how to Learn” aims to understand the promise and limitations of the in-context learning abilities of transformer models to meta-learn algorithms for strategic data collection. Personnel: PI: Aldo Pacchiano, PhD, Assistant Professor (CDS); Collaborating Investigators: Alessio Russo, Postdoctoral Associate (CDS); Students: Yilei Chen, PhD student, ENG (ECE); Aida Afshar, PhD Student, CDS; Yichen Song, PhD Student, CDS.
19) Project: “Enhancing Reasoning in Artificial Intelligence Models” aims to understand and enhance abstraction and analogical reasoning capabilities of artificial intelligence (AI) models. Specifically, this project attempts to create new methodologies for bootstrapping LLMs’ weak-to-strong generalization in reasoning tasks. Personnel: PIs:Aldo Pacchiano, PhD, Assistant Professor, CDS; Michael Hasselmo, PhD, Professor, CAS (Psych & Brain Sciences), ENG; Students: Yilei Chen ENG (ECE), Mark Tracy, ENG (SE).
20) Project: “Versatile, Safe and Personalizable Large Language Models” aims to advance methodologies for aligning language models by incorporating signals beyond pairwise preference data, prioritizing safety as a central optimization objective, and enabling rapid personalization to align outputs with specific user needs. Personnel: PI: Aldo Pacchiano, PhD, Assistant Professor (CDS); Collaborating Investigators: Alessio Russo, PhD, Postdoctoral Associate (CDS); Students: Yichen Song (CDS), Aida Afshar, PhD Student, CDS, Yilei Chen, PhD student, ENG (ECE).
21) Project: This research will design, develop, and make publicly available the models and automated tools to facilitate rapid image-based cellular analysis in basic biological research and other biotechnology systems. Personnel: PI: Bryan Plummer, PhD, Assistant Professor, CAS (CS).
22) Project: “BabyGPT: Developmentally Plausible Learning of Multimodal Foundation Models” will leverage multimodal self-supervised learning on an ego-centric video dataset recorded from the perspective of young children in order to model human-like one-shot learning and examine how language may facilitate vision-related tasks to advance both cognitive science and machine intelligence, contributing to the development of efficient, developmentally plausible AI systems. Personnel: PI: Venkatesh Saligrama, PhD, Professor, ENG (ECE, SE); Collaborating Investigator: Boqing Gong, PhD, Assistant Professor, CAS (CS).
23) Project: “Large Graph Models for Cybersecurity” engineers Knowledge Graph (KG) and Graph-based Retrieval-Augmented Generation (GraphRAG) methods to automate cybersecurity operations, such as root-cause analysis of vulnerabilities and detection of errors in threat databases. Personnel: PI: David Starobinski, PhD, Professor, Engineering (ECE, SE), CAS (CS).
24) Project: “Developing an AI-based Content Moderation Pipeline for Social Media” aims to analyze and process social media content, with the goal of identifying content that should receive moderation labels (e.g., sensitive content, hate speech, or false information) using Learning to Rank techniques, Large Language Models, and image processing techniques to create a multi-modal decision agent. Personnel: PI: Gianluca Stringhini, PhD, Associate Professor, ENG (ECE).
25) Project: “Analyzing Global Election Discourse” aims to conduct cross-national analyses of political narratives and information diffusion during the 2024 global election cycle. This project will examine interactions among citizens, politicians, media institutions, and automated actors, and explore how visual imagery shapes public perception. Personnel: PI: Chris Chao Su, PhD, Assistant Professor, COM (Emerging Media), CDS.
26) Project: “Developing a Neuropathology Foundation Model for Advancing Analysis of Neurodegenerative Diseases” aims to address existing barriers by developing the first neuropathology-specific foundation model, NeuroPath, leveraging the world’s largest digital neuropathology database from Boston University Alzheimer’s Disease Research Center (ADRC) Digital Pathology Core to fully unlock the potential of AI-based analysis in advancing our understanding of Neurodegenerative Diseases. Personnel: PIs: Lei Tian, PhD, Associate Professor, ENG (ECE, BME); Jonathan Cherry, PhD, Assistant Professor, Chobanian & Avedisian SOM (Pathology & Laboratory Medicine, Neurology and Anatomy & Neurobiology).