Researcher · Engineer · Speaker

Jaya Preethi MohanDoctoral Researcher & Computer Scientist

AI / ML·HPC·Cloud·Quantum·Power Systems

UNDResearch
PNNLNational Lab
IBMTechnology
About

Research at the Intersection

As an experienced Doctoral Researcher at the University of North Dakota, a former Computer Scientist at Pacific Northwest National Laboratory, and with prior industry experience at IBM, I have led research and engineering work spanning AI systems, cloud infrastructure, and national critical infrastructure across research and industry settings.

Applied contributions span production ML systems for power grid security, statistical anomaly detection for IT/OT environments, and agentic AI frameworks for distributed energy resource management, with peer-reviewed publications across IEEE venues.

In cloud and DevOps engineering, I design serverless AWS architectures, implement CI/CD pipelines synchronized with cybersecurity maturity assessments, and build containerized cross-environment workflows using Docker and Singularity that enable reproducible science across DOE supercomputers, cloud platforms, and institutional HPC clusters.

I create accessible technical content, contribute to tech communities, stay active through CrossFit, tennis, and golf. I value thoughtful conversations about ethics, optimism, meaningful use of time, and life beyond work.

AI Research Anomaly Detection & Algorithm Design HPC & Systems CUDA · SLURM · MPI Power Systems Research & Talks IEEE Publications Technical Talks & Workshops Cloud & DevOps AWS · Serverless CI/CD & Containerization
Research

Current Focus Areas

Investigating where machine learning meets physical systems, HPC infrastructure, and deployment reproducibility.

Doctoral Thesis · In Progress
Reproducibility Limits in Scientific Machine Learning
Demonstrating that containerized deployments fail to preserve scientific validity even when code, weights, and containers are identical. Introduces the Physical Plausibility Score (PPS) metric and identifies critical batch-size thresholds below which physics constraints degrade super-linearly.
Physics-Informed MLContainersReproducibilityPPS MetriccuDNN
Core Architecture
Physics-Informed Graph Attention Networks (PI-GAT)
A graph attention network for power grid anomaly detection on the IEEE 118-bus benchmark, integrating physical constraint layers that enforce power flow equations. Trained on UND's Talon HPC cluster with PyTorch DDP and Singularity containers.
Graph Neural NetworksPower SystemsPyTorch DDPIEEE 118-Bus
Emerging Direction
LLM Deployment & Agentic AI Orchestration
Scalable deployment pipelines for large language models and agentic AI orchestration frameworks for energy systems, bridging containerized ML infrastructure with cloud platforms, quantum computing foundations, and real-time decision support.
LLM WorkflowsAgentic AICloudQuantumEnergy Systems
Experience

Professional Journey

2024 – Present
Doctoral Researcher
University of North Dakota · Grand Forks, ND
Investigating physics-informed machine learning, quantum-classical hybrid computing, and LLM-driven agentic orchestration for power systems and distributed energy resources. Research addresses fundamental questions in reproducibility, physical constraint preservation, and intelligent grid automation across HPC and containerized environments.
2023 – 2026
Computer Scientist 2 · Power Systems Research
Pacific Northwest National Laboratory · Richland, WA
Led selected technical components and applied R&D across power grid analytics, cybersecurity infrastructure, and cloud ML deployment. Developed production-grade anomaly detection systems, DER interoperability frameworks using OpenFMB and MQTT, and CI/CD pipelines tied to automated cybersecurity maturity assessments. Contributed to DOE critical infrastructure initiatives including CEASER and the North American Energy Resilience Model.
2023
Masters Research Intern
Pacific Northwest National Laboratory · Richland, WA
Deployed the ExaGO optimization toolkit for exascale power grid simulations on the Frontier supercomputer. Built CI/CD pipelines and containerized workflows using Docker and Singularity for reproducible multi-site scientific computing across DOE facilities and cloud platforms.
2021 – 2023
Graduate Research Assistant
University of North Dakota · Grand Forks, ND
Conducted FAA-funded ML research on unmanned aerial systems safety, GPS anomaly detection, and electromagnetic interference mitigation. Built GPU-accelerated deep learning pipelines on the UND Talon HPC cluster using CUDA, SLURM, and distributed training frameworks. Published peer-reviewed research on GPS false data injection detection, 5G interference mitigation, and a hierarchical software framework for UAS national airspace integration.
2019 – 2021
Technical Support Associate & Service Delivery Specialist
IBM · Bangalore, India
Started as a Technical Support Associate resolving Red Hat Enterprise Linux and virtual machine issues, handling global outages and critical system incidents. Progressed to Service Delivery Specialist in Cloud Managed Backup, managing data recovery and disaster management operations using IBM Spectrum Protect, Veritas, and Veeam across IBM and enterprise client environments.
Expertise

Technical Proficiency

AI & Machine Learning
  • Physics-Informed Neural Networks
  • Graph Neural Networks (GAT)
  • Large Language Models
  • PyTorch · PyTorch DDP
  • Distributed Training
HPC & Cloud Infrastructure
  • CUDA · OpenMP · MPI
  • SLURM · HPC Clusters
  • Docker · Kubernetes · Singularity
  • AWS · Azure · GCP
  • Red Hat · OpenShift
Power Systems & Data
  • ExaGO · pandapower · PowerWorld
  • SCADA / OT Systems
  • IEEE Test Benchmarks
  • HDF5 · Zarr · FAIR Principles
  • Python · C++ · Bash
Recognition

Awards & Honors

Outstanding Performance Award
Pacific Northwest National Laboratory
Shining Star Award
IBM
Best Research Poster Award
IEEE CARS
Publications

Selected Works

NeurIPS 2026 · Under Review
On the Limits of Reproducibility in Scientific Machine Learning: When Containers Fail to Preserve Physical Constraints
J. P. Mohan, P. Ranganathan
IEEE Cloud Summit 2025
Containerized LLM Deployment Architectures for Scientific Computing
J. P. Mohan et al.  ·  doi: 10.1109/Cloud-Summit64795.2025.00027
View →
IEEE CCWC 2024  ·  pp. 198–206
Hierarchical Software Framework for Safe Unmanned Aerial Systems Integration into National Airspace (NAS)
J. P. Mohan, P. Ranganathan, H. Reza  ·  doi: 10.1109/CCWC60891.2024.10427776
View →
IEEE ESCS 2023 · Las Vegas, USA
False Data Injection (FDI) Modeling and Detection in Global Positioning Systems (GPS) for UAS Environments
J. P. Mohan, P. Ranganathan  ·  doi: 10.1109/ESCS.2023.10487453
View →
IEEE eIT 2022  ·  pp. 446–454
Cyber Security Threats for 5G Networks
J. P. Mohan, N. Sugunaraj, P. Ranganathan  ·  doi: 10.1109/eit53891.2022.9813965
View →
Writing

Research Blog

Notes on scientific AI infrastructure, HPC workflows, reproducibility in machine learning, and energy systems research.

Connect

Let's Collaborate

Exploring research scientist and data scientist roles at national laboratories and leading technology organizations. Open to conversations about research collaboration and opportunities.