CREDIF spring 2025 AI awards & fellowship recipients

Capital One established an Academic Center of Excellence with University of Southern California Viterbi School of Engineering and Computing in 2024. The Center for AI and Responsible Decision Making in Finance (CREDIF) has a focus on supporting innovative AI research.

Each year the center will host an annual call for proposals. This call invites faculty members to submit AI research proposals to be considered for a research award or to nominate a student to be the recipient of a Capital One Fellowship. The first call for proposals was held from August 2024 - October 2024. Award recipients began their projects in January of 2025. Their work is driving technological advancements in the financial services sector by enhancing risk management and decision-making.

The first call solicited proposals that were focused on research proposals in the areas of:

• Robust AI in the face of noisy data and labels

• Synthetic data generation with an emphasis on generation of structured/tabular data

• Explainability framework that improves transparency and interoperability in LLM based systems (particularly in the the presence of multiple agents) 

• Integration of graph knowledge representations with AI

After careful consideration, the advisory board selected six USC research proposals to receive AI research awards and six USC PhD students to receive AI fellowship awards. ​​Fellows and award winners participate in collaborative research initiatives to advance AI technologies in finance.

The Research and Fellowship award recipients will engage with Capital One research scientists throughout the year to provide industry insights and help to further inform their research projects. Our research scientists are committed to forming strong connections with the faculty and students, which includes meeting regularly on the research projects, providing industry insights and contributing to the overall success of the research. 

Below you can learn more about each of the AI award recipients and their focus areas for the year.

“Privacy preserving synthetic data generation with differentially private reward model” by Sai Praneeth Karimireddy, Assistant Professor in the Lord Thomas Department of Computer Science, and Robin Jia, Assistant Professor in the Lord Thomas Department of Computer Science and Lead of Allegro Lab. 

“Reliable AI predictions from imperfect financial datasets” by Matteo Sesia, Assistant Professor in the Department of Data Science and Operations at the Marshall School of Business.

“Disentangling and explaining model-level subjectivity with diverse large language models in multi-agent decision-making” by Shrikanth Narayanan, University Professor, Nikias Chair in Engineering and VP for Presidential Initiatives research. 

“Knowledge graph-enhanced RAG for financial genAI systems” by Viktor Prasanna, Charles Lee Powell Chair in Engineering, Professor of Electrical Engineering and Professor of Computer Science Director of the Center for Energy Informatics.

“Learning predictive models using imperfect data: an integrated continuous-discrete approach” by Andres Gomez, Assistant Professor of Industrial and Systems Engineering, and Johannes Royset, Professor of Industrial and Systems Engineering.

“Revolutionizing financial fraud detection: advanced graph-based anomaly detection with LLM integration” by Yue Zhao, Assistant Professor in the Lord Thomas Department of Computer Science and Jieyu Zhao, Assistant Professor in the Lord Thomas Department of Computer Science.

Yuvuz Faruk Bakman, Ph.D. student, is researching LLM trustworthiness by mitigating uncertainty and hallucinations.

Advised by: Salman Avestimehr

Yuvuz is currently focused on improving the trustworthiness and accuracy of large language models, exploring uncertainty and hallucinations, which has yielded an improved focus on continual learning, self-supervised contrastive learning and federated learning.

Akhil Agnihotri, Ph.D. student, is researching reinforcement learning to build safer AI systems using human feedback. 

Advised by: Rahul Jain

Akhil's research focuses on advancing safer and more responsible artificial intelligence through reinforcement learning—a type of AI that learns by interacting with its environment to make better decisions over time. His work addresses critical challenges in deploying AI systems for real-world applications such as robotics, autonomous vehicles and large-scale AI models. A key part of his research involves building algorithms that allow AI systems to operate under strict safety and efficiency constraints, ultimately ensuring responsible behavior even in complex or uncertain environments. To align AI systems with human values, Akhil has worked on preference-based learning approaches, culminating in algorithms which excel at leveraging human feedback to personalize AI behavior while maintaining strong theoretical guarantees. Akhil has published in top-tier machine learning conferences like ICML and NeurIPS and has used his experience in the financial services industry to collaborate with Google Deepmind, Qualcomm and most recently capital one during his Ph.D.

Kien Nguyen, Ph.D. student, is researching graph foundation models to develop robust and fair graph representations.

Advised by: Paul Bogdan

Kien’s research primarily focuses on Machine/Deep Learning and Graph Neural Networks. His published research has focused on learning robust, fair graph representations under constrained scenarios: data imbalance, data structural bias or type inaccessibility. Currently, he is working on developing graph foundation models (GFM), a single model/system able to generalize to new, unseen graph data across various tasks. Kien researched how LLM’s embed features into a unified semantic space, advanced graph techniques are investigated to capture contextual and topological patterns among graphs. With the growing popularity of graph modeling in diverse domains, effective and efficient GFM has huge potential for both scientific and industrial applications.

Jingwei Ji, Ph.D. student, is researching online learning to enhance decision-making under financial uncertainty.

Advised by: Jong-Shi Pang

Jingwei explores unique challenges in finance through the lens of online learning, enhancing the understanding of decision-making under uncertainty by providing both complexity insights and practical algorithmic solutions. During his Ph.D., he has had the privilege to be advised by Professor Renyuan Xu and Professor Jong-Shi Pang. Previously, he has also conducted research in approximate dynamic programming and queueing theory.

Jungie Ye, Ph.D. student, is researching scalable AI agents for interpretable decision-making using large-scale data and foundational models.

Advised by: Yue Wang

Jungie’s research aims to develop scalable AI agents for interpretable decision-making by harnessing large-scale data and foundational models. This work is grounded in computer vision and robotics, where Jungie focuses on creating generalizable AI systems capable of perceiving and interacting with diverse environments. Ultimately, his goal is to advance responsible AI by designing models that are both scalable and adaptable, facilitating trustworthy and effective deployment in real-world applications.

Priyanka Dey, Ph.D. student, is researching how LLMs capture personality traits to enable culturally aware AI personalization.

Advised by: Emilio Ferrara

Priyanka’s research focuses on understanding how large language models (LLMs) represent personality traits and cultural backgrounds, particularly through psychometric analysis and country-specific personas. She is interested in exploring the intersection of personality modeling, cultural diversity, and AI personalization, investigating how different datasets and prompting strategies influence LLMs' adaptation to human-like traits. Additionally, she is interested in leveraging this knowledge to generate more personalized content by designing templatized and customized prompts, ensuring AI systems can better align with users' preferences and cultural contexts.

Research award recipients kicked off Spring 2025 projects in January and will continue through December. Additionally, the call for proposals for the Fall 2025 - Spring 2026 went out to all Viterbi School of Engineering and Computing Faculty at the end of January. In the coming months, CREDIF will announce future winners, recognizing innovators driving breakthroughs in machine learning and AI research.

Through CREDIF and other partnerships, we're eager to tackle some of the most challenging industry problems at the intersection of AI, finance and banking. Our achievements include developing AI-powered solutions that analyze complex financial data, apply advanced analytics and drive transformative applications. Learn more about our recent industry collaborations and more information on our academic partnerships.

New to tech at Capital One? We're building innovative AI solutions in-house and transforming the financial industry:

• Explore our AI research: Dive deep into our latest advancements in AI and machine learning.

• Discover career opportunities: Learn about exciting applied research career paths at Capital One for researchers and engineers passionate about AI and join our world-class team.

• Engage with our team: Meet our researchers and AI experts, ask questions and discuss the future of AI in finance.