Computer Engineering Commons^™

A Hands-On Lab On Agentic Scientific Discovery For Time Series With The Model Context Protocol, Dhaval Patel, Miaomiao Feng, Chathurangi Shyalika, Shuxin Lin, Hasnat Md. Abdullah

Publications

Agentic AI offers new opportunities to accelerate scientific discovery for time series by enabling autonomous systems to inspect data, formulate hypotheses, compose experiments, and refine models based on empirical evidence. However, building agents that make sound modeling decisions requires more than connecting a language model to forecasting tools. Key challenges include exposing large model catalogues without overwhelming the agent, grounding parameter choices in data characteristics, supporting iterative experimentation, and evaluating the quality of the reasoning behind each decision.

This hands-on lab introduces participants to agentic scientific discovery for time series using the Model Context Protocol (MCP). Through five interactive sprints, …

Securing Reconfigurable Systems: Real-Time Ml-Based Hardware Trojan Detection Via Binary Fpga Bitstream Analysis, Rye C. Stahle-Smith Mr.

Caroliniana Undergraduate Research Journal

Reconfigurable computing platforms such as Field-Programmable Gate Arrays (FPGAs) are widely deployed in high-performance embedded systems and AI acceleration workloads. Their post-deployment reconfigurability, however, introduces a critical attack surface: maliciously crafted configuration bitstreams can embed hardware Trojans capable of leaking sensitive data, degrading system reliability, or silently altering computation. This paper presents an AI-based predictive modeling framework for detecting malicious FPGA configurations directly from binary bitstreams, without reliance on source code or netlists. Bitstreams sourced from state-of-the-art (SOTA) Trust-Hub benchmarks were re-engineered for the Xilinx PYNQ-Z1 platform and labeled as benign, malicious, or empty. A 256-dimensional byte-frequency feature vector is …

Context Matters: Evaluating Llm-Generated Knowledge Graph Schemas, Ritvik Garimella, Riju Marwah, Atishay Jain, Khusham Bansal, Amit Sheth

Publications

Knowledge graph (KG) schema engineering is labor-intensive and resists automation at scale. We investigate whether LLMs can generate domain-specific KG schemas of sufficient quality for downstream symbolic reasoning. We propose a tiered contextual framework that varies domain context richness across four levels: zero context, domain scope, task requirements, and data distribution. Generated schemas are evaluated intrinsically on BioRED (600 PubMed abstracts, multi-type entities and relations), where automated tiered schemas match an established KG construction baseline at 79.9% EC, with edge conformance rising from 47.5% at L1 to a stable 78–80% from L2 onward. Extrinsic evaluation on a 50-record MedHop controlled …

Doing Less With More: A First-Principles Exploration Of The Suitability Of Agentic Computing Over Alternative Architectural Choices, Ritvik Garimella, Biplav Srivastava, Amit Sheth

Publications

There is growing interest in automating business activities with Agentic Artificial Intelligence (AI) due to latter's seeming ease of use. Never has it been easier, or costlier, to do less with more. However, little is known about when agents are preferable to established alternatives such as local computation, Representational State Transfer (REST), the Simple Object Access Protocol (SOAP), and the Model Context Protocol (MCP), particularly when development speed, performance, and operational cost are considered. We investigate this question using a controlled mathematical task that compares seven methods on a benchmark of 1,000 arithmetic expressions where semantics of operator precedence has …

One Size Does Not Fit All: Revisiting World Models And Neurosymbolic Ai, Amit P. Sheth, Madhur Thareja, Anushka Pawar, Niyati Rawal

Publications

World models are being built twice, from opposite ends, without a shared theory of how the two halves should meet. One lineage grounds the world model in perception: a self-supervised, latent-predictive encoder – exemplified by Joint Embedding Predictive Architectures (JEPA) – that learns the structure of sensory experi-ence. A second, older lineage grounds the world model in cognition: an explicit, inspectable structure of entities, rules, and constraints, ranging from knowledge graphs to formal logic to physical law. Neither lineage alone has produced a world model that is simultane-ously adaptive and auditable. We argue this is not solved by picking a …

Rgfc: Registry Grounded Function Calling With Stage-Decomposed Evaluation, Ritvik Garimella, Vedant Khandelwal, Anvi Kohli, Amit Sheth

Publications

Tool-calling evaluations often report whether the final call matches a reference answer, but this single outcome merges different failures: choosing a function outside the available registry, violating the selected schema, or filling arguments with values not supported by the user request. These failures matter more when models must choose from larger tool registries. We introduce RFCD, a registry-scaled function-calling benchmark built from BFCL, Glaive Function Calling v2, and Tool-Call-Data, with standardized JSON-schema registries ranging from 5 to 3,000 functions. We also introduce TAAG, a deterministic stage-decomposed evaluator for registry conformance, structural completeness, and argument grounding. Across six locally served sub-4B …

Advancing Edge Ai Through Integrated Neuromorphic Algorithms And Hardware, Peyton Chandarana

Theses and Dissertations

As machine learning continues to expand across diverse sectors, including healthcare, transportation, environmental monitoring, and consumer electronics, many applications operate under strict constraints on latency, bandwidth, privacy, and power availability. Meeting these demands increasingly requires deploying machine learning at the edge to process data locally rather than relying on remote cloud infrastructure. These limitations highlight the importance of computational efficiency and motivate the exploration of alternative hardware and algorithmic paradigms capable of delivering real-time, accurate inference under tight resource constraints.

Two major paradigms have emerged for edge intelligence: conventional artificial neural networks (ANNs) deployed on highly parallel edge accelerators such …

Generalized Planning Using Language Models And Its Applications, Vishal Pallagani

Theses and Dissertations

Planning is a fundamental capability for intelligent systems, yet classical planners remain difcult to scale and generalize across domains due to their reliance on hand- engineered models, brittle search procedures, and limited adaptability. These limita- tions constrain real-world applications that require fexible reasoning, robust decision- making, and the ability to operate across diverse environments.

Large language models have emerged as powerful learners supported by their ar- chitectures and large-scale training data, demonstrating strong capabilities in tasks such as question answering, code generation, summarization, and reasoning. Moti- vated by their broad generalization abilities, I investigate whether language models can be systematically …

Biologically-Inspired Multiscale Neuromorphic Architecture, Christian O'Reilly, Ramtin Zand

Publications

This white paper proposes a biologically-inspired multiscale neuromorphic architecture that bridges key gaps between artificial neural networks (ANNs), spiking neural networks (SNNs), and biological neural networks (BNNs). While SNNs offer promising energy efficiency, their broader adoption remains limited by suboptimal performance and the need for novel learning paradigms. To address these challenges, the proposed framework integrates structural and functional principles observed in the brain, including hierarchical organization, sparse and modular connectivity, predictive coding, and diverse neuronal dynamics.

The architecture operates across micro-, meso-, and macro-scales, incorporating neuron-level diversity (e.g., excitatory/inhibitory and principal/support cells), canonical microcircuits (CMCs), and large-scale hierarchical organization. …

In-Situ Eval: A Modular Framework For Custom And Real-Time Rag Benchmarking, Ritvik Garimella, Kaushik Roy, Chathurangi Shyalika, Amit Sheth

Publications

Retrieval-Augmented Generation (RAG) has become the standard approach for integrating domain knowledge into Large Language Models (LLMs). However, fair comparison of RAG pipelines remains difficult: data preparation is often ad hoc, subsampling methods are opaque, parameters vary across implementations, and evaluation is fragmented. We present In-Situ Eval, a unified and reproducible framework that operationalizes the full RAG pipeline with configurable subsampling strategies and both RAG-specific and generic evaluation metrics. The platform supports two execution modes: an offline Dataset mode for evaluating precomputed outputs, and a live Retrieval mode for benchmarking RAG variants with state-of-the-art LLMs. Users can flexibly select datasets, …

When Disasters Trigger Cyber Vulnerabilities: Mapping Physical-Digital Interdependencies In Critical Infrastructure Systems, Dikshya Panta, Sicheng Wang, Aditya Sapkota, Prakash Ranganathan

Faculty Publications

Critical infrastructure (CI) systems such as power, water, communications, and emergency services are increasingly exposed to compound risks in which natural disasters and cyber incidents interact and amplify one another. Traditional risk assessments often isolate physical and digital threats, overlooking the cascading dependencies that emerge when operational stress, emergency reconfiguration, and adversarial exploitation coincide. This study conducts a 2019–2025 scoping review and introduces a Geographic Information System (GIS) driven six-stage disaster cyber compounding framework that characterizes, maps, and operationalizes compound risk across interdependent CI sectors. The framework integrates a common operating picture, analytic situational understanding, exposure mapping, threat-fingerprint encoding, detection …

Toward Neurosymbolic Reinforcement Learning Via Editable Specifications, Vedant Khandelwal, Hong Yung Yip, Amit Sheth

Publications

Reinforcement learning systems are commonly adapted to new settings by retraining or fine-tuning policies. This default is costly, difficult to audit, and poorly aligned with structured requirement changes such as revised safety rules, new operational constraints, or updated user preferences. We argue for an alternative abstraction: adaptation via edits to an external, human-readable specification that the agent consults at execution time. We propose conditioning decision-making on an editable knowledge graph encoding (i) rules capturing action applicability and high-level effects, (ii) hard constraints defining feasibility, and (iii) soft preferences shaping tradeoffs among feasible behaviors. Requirement changes become graph edits, not policy …

Crisis Observatory: Extracting Credible Signals During A Crisis In The Age Of Llms, Kuan-Chieh Lo, Pranav Maneriker, Sriram Sai Ganesh, Dominik Winecki, Kelly Garrett, Ayaz Hyder, Arnab Nandi, Valerie Shalin, Shannon A. Bowen Ph.D., Amit Sheth, Srinivasan Parthasarathy

Publications

Systems for crisis response have required several different models for the analysis of unstructured text, such as identifying needs, locations, topics, routing, and matching of needs with available responders. Large Language Models (LLMs) have replaced task-specific models across various language processing tasks. However, LLMs are known to be limited by their training data, collected before the crisis. In this demo, we explore the use of LLMs for crisis response scenarios with rapidly evolving information environments. We show how the augmentation of these models with external reliable sources of crisis-specific information can help build adaptive systems for response. The demonstration video …

Building Multimodal Knowledge Graphs: Automation For Enterprise Integration, Ritvik Garimella, Hong Yung Yip, Revathy Venkataramanan, Amit P. Sheth

Publications

As enterprises increasingly aim to incorporate artificial intelligence into their workflows to tackle complex, multimodal tasks, the demand for intelligent, robust, and trustworthy systems is paramount. While multimodal large language models offer initial capabilities for processing diverse data streams, their dependence on embedding-based representations limit their effectiveness in delivering semantically grounded explanations and reasoning, as well as qualities essential for enterprise-grade applications. Neurosymbolic approaches provide a promising alternative by enabling traceable, context-aware decision making. However, constructing enterprise-level multimodal knowledge graphs (MMKGs) that enable neurosymbolic approaches remains largely impractical. Although prior efforts have explored MMKG construction, they fall short in addressing …

Enhancing Visibility On A Science Dmz With P4-Perfsonar, Ali Mazloum, Elie Kfoury, Ali Alsabeh, Jose Gomez, Jorge Crichigno

Faculty Publications

The Science Demilitarized Zone (Science DMZ) is a specialized network designed to facilitate the transfer of large-scale scientific data. One of the key elements of the Science DMZ is perfSONAR, an active performance measurement device that monitors end-to-end paths over multiple domains. Although versatile, perfSONAR faces limitations such as restricted visibility of events and coarse-grained measurements. This paper proposes a scheme that integrates P4 programmable data plane (PDP) switches with perfSONAR. P4 PDP switches are passively installed and operate on real-time traffic copies, providing flexibility to collect fine-grained custom measurements and report events in the data plane. This integration enables …

Hardware Accelerated Simulation Of Buck Converters Using Physics-Informed Neural Networks, James Clayton Crews

Theses and Dissertations

Physics-informed neural networks (PINNs) are an emerging machine learning method for learning the behavior of physical systems described by governing differential equations. Dc-dc power-electronic converters are used in a variety of industry applications such as motor drives or power supplies where real-time simulation is critical for control and safety. This thesis investigates physics-informed machine learning as an approach to develop a real-time digital twin for dc-dc power converters. Traditional numerical integration methods are used to approximate discretized behavior, and the results are compared with a trained PINN model. Modern ML frameworks (such as PyTorch and TensorFlow/Keras) are used to quickly …

From Assembly Lines To The Open Road: Predicting Rare Events In Autonomous Systems, Ruwan Wickramarachchi

Publications

In the age of embodied AI and smart automation, autonomous agents are increasingly deployed in high-stakes, real-world environments. Ensuring the robustness and resilience of these systems in the face of rare but critical failure events is essential for their safe and reliable operation. Accurate forecasting of such rare events is particularly crucial, as a single overlooked anomaly can lead to catastrophic consequences. In manufacturing, for instance, unplanned downtime due to rare failures costs industries over \$50 billion annually, with sectors like automotive losing more than \$2 million per hour—even with preventive maintenance systems in place.

However, the extreme rarity and …

Multiagent Copilot In Industrial Ai Applications, Chathurangi Shyalika, Renjith Prasad, Utkarshani Jaimini, Cory Henson, Fadi El Kalach, Amit Sheth

Publications

In the era of smart automation and digital transformation, achieving efficiency, precision, and adaptability is essential for industries to remain competitive. Sectors, including manufacturing, supply chain and logistics, healthcare, finance, and retail, face significant challenges in deploying Artificial Intelligence (AI) solutions tailored to their unique needs, particularly in critical, resource-constrained applications. According to Gartner’s 2024 Hype Cycle for Artificial Intelligence, composite AI, which integrates techniques like machine learning, knowledge graphs, and rule-based systems, is becoming foundational for industries, enhancing predictions, decisions, and scalability across complex environments.

The complexity of real-world systems requires Industrial AI solutions to be customizable to business …

Swvl: A Custom Ai-Powered Face Tracking Camera Gimbal, Alexander J. Anderson-Mcleod, Jakub Jerzmanowski, Michael Laitarovsky, Trevor Allison, Jagger Tanner

Senior Theses

In response to the growing demand for smarter, more responsive face tracking cameras in the post-pandemic world, our team designed SWVL, a custom AI-powered face tracking gimbal meant to address the limitations commonly encountered by the commercial models currently on the market. These commercially available gimbals come with several issues, such as frequently losing track of the person in the frame and requiring manual resets, which we sought to fix with our implementation. We designed a system with fully custom hardware and software including a 3D printed dual-axis camera gimbal driven by stepper motors, a control PCB based around an …

Cross-Layer Design And Optimization Of Analog In-Memory Computing Systems, Md Hasibul Amin

Theses and Dissertations

There has been a rapid growth in the computational demands of machine learning (ML) workloads in recent days. Conventional von Neumann architectures are not capable of keeping up with the high cost of data movement between the processor and memory, well-known as memory wall problem. In-memory computing (IMC) has been focused as a solution by the researchers, where the computation is performed inside the memory devices such as SRAM, MRAM, RRAM etc. Most commonly, the memory devices are arranged in a crossbar setting where the matrixvector multiplication (MVM) operation is performed through intrinsic parallelism of analog computations. The conventional IMC …

Neurosymbolic Knowledge-Grounded Planning And Reasoning In Ai Systems, Amit Sheth, Vedant Khandelwal, Kaushik Roy, Vishal Pallagani, Megha Chakraborty

Faculty Publications

To build AI systems capable of decision-support assistance, such as AI-assisted healthcare, it is essential to develop user-centric decision-making processes that are robust, interpretable, and capable of effectively processing and acting on natural language interactions. Instruction-based prompting of large language models has demonstrated considerable success in supporting humans with information assistance tasks, including creative writing and content generation. However, recent studies reveal that language models exhibit limitations in performing complex reasoning and planning tasks, such as constructing compositional or hierarchical plans involving multiple reasoning steps. To address these challenges, we propose a neurosymbolic framework that integrates large language models with …

Msbzip55 Regulates Salinity Tolerance By Modulating Melatonin Biosynthesis In Alfalfa, Tingting Wang, Jiaqi Yang, Jiamin Cao, Qi Zhang, Huayue Liu, Peng Li, Yizhi Huang, Wenwu Qian, Xiaojing Bi, Hui Wang, Yunwei Zhang

Faculty Publications

No abstract provided.

Lab: Developing Explainable Multimodal Ai Models With Hands-On Lab On The Life-Cycle Of Rare Event Prediction In Manufacturing, Chathurangi Shyalika, Ruwan Wickramarachchi, Revathy Venkataramanan, Dhaval Patel, Amit Sheth

Faculty Publications

In the age of Industry 4.0 and smart automation, unplanned downtime is costing industries over $50 billion annually. Even with preventive maintenance, industries like automotive lose more than $2 million per hour due to downtime caused by unexpected or "rare'' events. The extreme rarity of these events makes their detection and prediction a significant challenge for AI practitioners. Factors such as the lack of high-quality data, methodological gaps in the literature, and limited practical experience with multimodal data exacerbate the difficulty of rare event detection and prediction. This lab will provide hands-on experience to learn how to address these challenges …

A Survey On Food Ingredient Substitutions, Hyunwook Kim, Revathy Venkataramanan, Amit P. Sheth

Publications

Diet plays a crucial role in managing chronic conditions and overall well-being. As people become more selective about their food choices, finding recipes that meet dietary needs is important. Ingredient substitution is key to adapting recipes for dietary restrictions, allergies, and availability constraints. However, identifying suitable substitutions is challenging as it requires analyzing the flavor, functionality, and health suitability of ingredients. With the advancement of AI, researchers have explored computational approaches to address ingredient substitution. This survey paper provides a comprehensive overview of the research in this area, focusing on five key aspects: (i) datasets and data sources used to …

Exploring The Potential Of Large Language Models For Assisting With Mental Health Diagnostic Assessments: The Depression And Anxiety Case, Kaushik Roy, Harshul Surana, Darssan Eswaramoorthi, Yuxin Zi, Vedant Palit, Ritvik Garimella, Amit Sheth

Publications

Large language models (LLMs) are increasingly attracting the attention of healthcare professionals for their potential to assist in diagnostic assessments, which could alleviate the strain on the healthcare system caused by a high patient load and a shortage of providers. For LLMs to be effective in supporting diagnostic assessments, it is essential that they closely replicate the standard diagnostic procedures used by clinicians. In this paper, we specifically examine the diagnostic assessment processes described in the Patient Health Questionnaire-9 (PHQ-9) for major depressive disorder (MDD) and the Generalized Anxiety Disorder-7 (GAD-7) questionnaire for generalized anxiety disorder (GAD). We investigate various …

C 3 An: Custom, Compact And Composite Ai Systems - A Neurosymbolic Approach: 4Th-Generation Evolution Of Intelligent Systems, Amit P. Sheth, Kaushik Roy, Revathy Venkataramanan, Venkatesan Nadimuthu

Publications

Artificial Intelligence (AI) systems continue to evolve rapidly. From the architecture perspective, it is evolving from large, monolithic models trained on massive internet data to complex, multi-component “compound” systems and “agentic” frameworks capable of semi-autonomous decision-making. These systems show immense promise yet face numerous challenges in reliability, consistency, transparency, and alignment with user goals. In this article, we propose Custom, Compact and Composite AI with Neurosymbolic (C3AN) approach, a framework that paves way to 4th-generation of AI that integrates data, knowledge, and human expertise to build robust, intelligent and trustworthy AI systems defined by 14 foundation elements.

Custom emphasizes …

Safe Data-Enabled Control Of Human-In-The-Loop Robotic Manipulator Systems, Ritirupa Dey, Avimanyu Sahoo, Vignesh Narayanan

Publications

Safe control of human-in-the-loop (HIL) robotic manipulators is critical for applications such as assistive robotics, teleoperation in hazardous environments, and collaborative manufacturing. However, this remains challenging due to the lack of a unified framework that simultaneously addresses safety constraints, external disturbances, unmodeled dynamics, and dynamic role switching in the HIL setting. In this paper, we propose a novel NN-driven HIL control framework in which human–robot dyadic interaction occurs through the haptic channel. Using Lyapunov stability analysis, we theoretically show that the proposed NN-based controller ensures accurate joint trajectory tracking, compensates for system uncertainties, and adapts to human inputs modeled as …

Towards Rare Event And Anomaly Prediction In Manufacturing: Bridging Methodological Gaps In Industrial Applications, Chathurangi Shyalika, Renjith Prasad, Ruwan Wickramarachchi, Amit Sheth

Publications

Rare event prediction is critical in industrial applications, including real-world Industry 4.0 applications. These events, defined by their low occurrence frequency, are often difficult to predict due to the skewed data distribution, which complicates modeling and evaluation. In our research, we provide a comprehensive review of current approaches to rare event prediction across four key dimensions: rare event data, data processing techniques, algorithmic approaches, and evaluation methodologies [1]. By analyzing diverse datasets with multiple modalities, including numerical, image, text, and audio, we categorize the primary challenges and present the gaps in current research. Specifically, we present three novel research contributions …

Predicting Chaotic Systems With Quantum Echo-State Networks, Erik Connerty, Ethan N. Evans, Gerasimos Angelatos, Vignesh Narayanan

Publications

Recent advancements in artificial neural networks have enabled impressive tasks on classical computers, but they demand significant computational resources. While quantum computing offers potential beyond classical systems, the advantages of quantum neural networks (QNNs) remain largely unexplored. In this work, we present and examine a quantum circuit (QC) that implements and aims to improve upon the classical echo-state network (ESN), a type of reservoir-based recurrent neural networks (RNNs), using quantum computers. Typically, ESNs consist of an extremely large reservoir that learns high-dimensional embeddings, enabling prediction of complex system trajectories. Quantum echo-state networks (QESNs) aim to reduce this need for prohibitively …

A Benchmark Knowledge Graph Of Driving Scenes For Knowledge Completion Tasks, Ruwan Wickramarachchi, Cory Henson, Amit Sheth

Publications

Knowledge graph completion (KGC) is a problem of significant importance due to the inherent incompleteness in knowledge graphs (KGs). The current approaches for KGC using link prediction (LP) mostly rely on a common set of benchmark datasets that are quite different from real-world industrial KGs. Therefore, the adaptability of current LP methods for real-world KGs and domain-specific ap- plications is questionable. To support the evaluation of current and future LP and KGC methods for industrial KGs, we introduce DSceneKG, a suite of real-world driving scene knowledge graphs that are currently being used across various industrial applications. The DSceneKG is publicly …