Research · Oak Ridge National Laboratory
I work on making AI useful for science—and reliable enough for the decisions around it.
I’m a Postdoctoral Research Associate at Oak Ridge National Laboratory. My work brings together scientific discovery, agentic AI, and operational safety.
Some of the questions I study: Can AI help us evaluate a scientific hypothesis? Can a conversation between agents improve hazard analysis? What happens when the hardware behind a model goes wrong?
My background is in AI and hardware reliability, with a Ph.D. from UT Dallas. Outside the lab, I enjoy photography and sharing what I learn.
Degrees, institutions, advisors, and thesis research.
The University of Texas at Dallas · Richardson, Texas, USA
Advisor: Dr. Kanad Basu
Thesis: Functional Safety of Mission-Critical Hardware
Research on the functional safety of deep learning models and accelerators.
North Dakota State University · Fargo, North Dakota, USA
Advisor: Dr. Sumitha George
Thesis: Exploration of Multi-Threshold Ferro-Electric Field Effect Transistor Based Designs
Research appointments and industry experience.
Oak Ridge National Laboratory · UT-Battelle · Oak Ridge, Tennessee, USA
Enhancing Operational Safety via Agentic Dialogue Hazard Identification Analysis (SIGDIAL 2026).
Developing agentic workflows for scientific hypothesis generation, verification, and feedback.
TIES Research Lab · The University of Texas at Dallas
Investigated large language model vulnerabilities to adversarial bit flips (TMLR 2025).
Studied fault mitigation in deep neural networks using uncertainty quantification and multi-exit networks (DAC 2024).
Developed functional test patterns for detecting adversarial attacks on deep neural networks (HOST 2024).
Investigated binary neural network vulnerabilities to adversarial bit flips (TMLR 2024).
Developed graph-learning methods for fault criticality analysis in electrical and electronic systems (DAC 2024, DAC 2025).
Oak Ridge National Laboratory, Oak Ridge, TN, USA.
Developed a human-in-the-loop workflow using generative AI and reasoning language models for high-risk property identification.
Built an agentic workflow to study hazardous events at Department of Energy sites using historical incidents and risk analysis.
North Dakota State University · Fargo, North Dakota, USA
Designed a binary converter using ferroelectric field-effect transistors (FeFETs) and studied binary and ternary in-memory computing algorithms using QAHE memories.
Tata Consultancy Services · Bhubaneswar, India
Developed IFS application customizations using C#, Azure, PL/SQL, and Oracle Database.
Connecting scientific reasoning, operational safety, and the systems underneath AI.
Developing structured, reproducible workflows that help researchers generate and evaluate testable hypotheses. Current questions include knowledge-graph context, intrinsic model scoring, and scientific reasoning.
Investigating adversarial bit flips, hardware-induced faults, and fault mitigation. The work connects model behavior with its computing substrate, from binary networks and LLMs to compute-in-memory accelerators.
Figures reproduced from the original papers. Follow the project links for methods, results, and full author lists.
Selected papers & preprints · Full list on Scholar ↗
Guest lectures, graduate mentorship, and undergraduate laboratory instruction.
Delivered a graduate-level lecture on foundational concepts of deep learning, convolutional neural networks, recurrent neural networks, and transformers.
Mentored graduate research in circuit fault analysis.
Guided undergraduate students in developing strong foundations in programming. Emphasized clean coding practices, debugging strategies. Provided hands-on mentoring throughout implementation projects.
Guided undergraduate students in understanding digital systems, and structured design methodology. Discussed verification techniques, and disciplined engineering workflows. Provided hands-on mentoring throughout multi-stage design and implementation.
Awards, fellowships, and research recognition
AI-Powered cosmological simulations for dark energy signatures.
For “Adaptable Multi-L Voltage to Binary Converter Using Ferroelectric FETs,” with Arun Govindankutty, Shan Deng, Kai Ni, and Sumitha George.
Recent papers, awards, talks, and service.
16 updates
Received the ORPA All-Around Excellence Award.
Received the ORPEX President’s Award.
New preprint: “Rethinking Domain Specialization for Open-Ended Scientific Reasoning in Astronomy Language Models.”
Read preprint ↗New preprint: “Do LLMs Know a Good Hypothesis When They See One?” The study compares intrinsic model scoring with prompted hypothesis ranking.
Read preprint ↗Selected participant in the SC26 Early Career Program.
New preprint: “Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation.”
Read preprint ↗Paper accepted at Visual Analytics in the Age of Autonomous Science, held with IEEE VIS 2026.
Presented at the CCSD Summer Intern Summit, Oak Ridge National Laboratory.
ORNL profile ↗“Accelerating Compact UWB Fractal Antenna Design Using Interpretable Machine Learning with Forward and Inverse Prediction” accepted in Results in Engineering.
Paper ↗Research on human-guided hazard forecasting using work information and historical incidents.
Read paper ↗Research on retrieval-augmented, human-guided high-risk property classification.
Read paper ↗Reviewed papers for IEEE Design & Test and IEEE Transactions on Very Large Scale Integration Systems.
“Enhancing Operational Safety via Agentic Dialogue Hazard Identification Analysis” is available on arXiv.
Read preprint ↗New preprint: “The Compressive Knowledge Graph Hypothesis,” examining compact graph context for battery-materials hypothesis generation.
Read preprint ↗“Enhancing Operational Safety via Agentic Dialogue Hazard Identification Analysis” accepted at SIGDIAL 2026.
Read paper ↗Awarded for “AI-Powered cosmological simulations for dark energy signatures.” The symposium took place on May 6–7, 2026.
Research community service, volunteer organizing, and educational outreach.
Oak Ridge National Laboratory (ORNL) Postdoctoral Association Research Symposium 2026
Reviewer for IJCNLP-AACL, ASP-DAC, SOCC, ICPADS, ICCAD, VTS, DAC, DCAS, CVPR, IEEE Design & Test, and EACL.
Volunteered as a tutor in mathematics, physical sciences, and English, supporting students from rural backgrounds in building strong academic foundations and improving access to STEM education.
Graduate mentorship in circuit fault analysis through CE 6303: Testing & Testable Design.
Photography and writing outside the lab. Open any photograph to see the original.
Visitor traffic
Explore where visitors come from and how activity changes over time.
Loading visitor map…
Loading globe…
