Research Experience

Clemson University, Clemson, SC, US

• Design and conduct a large-scale survey study examining user perceptions of reactions to LLM data practices, with direct implications for AI system usability and trustworthiness.
• Implement experimental manipulations (e.g., anonymization, consent, model training) to evaluate their influence on participant decision-making and system adoption.
• Perform statistical analyses, including logistic mixed-effects modeling, post-hoc contrasts, and model fit evaluations, uncovering predictors of trust, safety, and comfort in AI contexts.
• Translate empirical findings into actionable design insights guiding the development of transparent, user-centered, and privacy-preserving AI systems.

• Assessed LLMs for their ability to process cultural nuances, contributing insights into fairness and inclusivity in AI interactions—awarded Best Short Paper at UMAP '25.
• Created culturally diverse scenarios to rigorously evaluate model accuracy and bias, informing potential strategies for LLM refinement.
• Developed Python pipelines to automate large-scale structured interactions with the ChatGPT API, enabling efficient high-volume data collection and analysis.

• Conducted NSF-funded research on privacy-preserving methods for user-centered AI studies.
• Analyzed data collection practices in the AI research community to identify risks to participant privacy and data integrity.
• Investigated technical approaches (e.g., differential privacy) to reduce risks of data tracing and improve responsible data use in AI-driven studies.

Zeva Incorporated, Fairfax, VA, US - Remote

• Studied security needs of three critical infrastructure sectors (including Financial Services Sector).
• Proposed countermeasures against potential threats, utilizing Zeva products.
• Wrote requirements for PKI frameworks in accordance with NIST, ANSI, and ISO standards.
• Specified how security methods (e.g., code signing) must be implemented.

Karlsruhe Institute of Technology, Karlsruhe, Germany

• Investigated voltage-based side-channel attacks in multi-tenant cloud FPGAs.
• Deployed cross-die sensors to analyze power fluctuations and identify potential attack vectors.
• Designed a real-time countermeasure to isolate malicious tenant regions—published in ICCAD ‘21.

Fraunhofer Institute for Experimental Software Engineering, Kaiserslautern, Germany

• Led the creation of a framework supporting Fraunhofer clients in safety-security analyses.
• Coordinated two work packages in the EU project SECREDAS among eleven European partners and authored comprehensive reports for the steering committee.
• Integrated digital signatures into safety-critical sensory data exchanged between autonomous vehicles to ensure data integrity.
• Enhanced the IEC 62559 template to improve usability for deriving nonfunctional requirements (e.g., security) for risk analysis teams in the energy industry.

SAP Research, Karlsruhe, Germany

• Designed and implemented a privacy-preserving benchmarking protocol enabling SAP clients to confidentially compare key performance indicators.
• Adapted ElGamal homomorphic encryption to support statistical computations (e.g. median).
• Sketched a formal security proof of the protocol.
• Analyzed protocol efficiency, demonstrating substantial performance improvements over existing solutions while maintaining equivalent security guarantees.

Siemens Digital Industries Software, Munich, Germany

• Designed an IoT prototype that detects and prevents pipeline leakage.
• Programmed NodeMCU platform to control water pumps based on inputs from sensors.
• Implemented a CoAP-based service to manage system components remotely.