Advancing Knowledge

Exploring the frontiers of AI and design to create technology that benefits humanity. Our research focuses on responsible AI development and human-centered design.

AI in MathArtificial IntelligenceMachine Learning

Can AI Invent Calculus, or Just Mimic Math?

Gauri Kale, Holly Diamond, Ava Hedayatipour*, Rahul Vishwakarma, Amin Rezaei

Presented at the 2026 International Conference on Advances in Artificial Intelligence and Machine Learning in Tokyo, Japan.

AAIML 2026
Deep Residual LearningHyper-ConnectionsStable & Efficient

Sparse Selective Hyper-Connections: A Unified Framework for Stable and Efficient Deep Residual Learning

Shrey Modi, Rahul Vishwakarma, Renju Rajeev, Tanay Parikh, Krishna Patel and Holly Diamond

Presented at SoutheastCon 2026 in Huntsville, AL, USA.

SoutheastCon 2026
Research IntegrityImpact ScoresSelf-Citations

Poster: Where Do Billions in Research Funding Really Go? When Self-Citations Inflate Impact Scores by 20%

Rahul Vishwakarma (WorkOnward) & Sinchan Banerjee (UST GlobalInc)

Poster presentation at PyTorch Conference 2025 in San Francisco.

PyTorch Conf 2025
Digital TwinStorage DevicesAI Architecture

Design Specification and AI-Driven Digital Twin Architecture for Storage Devices

Rahul Vishwakarma (WorkOnward) & Hemant Gaikwad (Dell Technologies)

Presented at SNIA Storage Developer Conference 2025.

SDC USA 2025
Autonomous Agentic AIData EngineeringGenAI

AAA-IDE: Autonomous Agentic AI for Data Engineering

Rahul Vishwakarma

Presented at the IEEE Enterprise Generative AI Summit 2025 in San Jose, CA.

IEEE GenAI 2025
Safety GuardrailsEfficiencyOrchestration

SafeRoute: Adaptive Model Selection for Efficient and Accurate Safety Guardrails in Large Language Models

Seanie Lee*, Dong Bok Lee*, Dominik Wagner, Minki Kang, Haebin Seong, Tobias Bocklet, Juho Lee, Sung Ju Hwang

An adaptive framework for selecting safety guardrails to balance efficiency and accuracy in LLM deployment.

ACL 2025 Findings
Multi-Agent AISignaling PrimitivesRobust AI

Endocrine-to-Synaptic: Learnable Signaling Primitives for Robust Multi-Agent AI

Rahul Vishwakarma + collaborators from USC & Emory

Presented at the AI Transformation in Industries and Infrastructure Track.

IEEE New Era AI 2025
Agent ProtocolBCA2PBio-Causal Framework

A Novel Bio-Causal Agent-to-Agent Protocol (BCA2P) Framework

Rahul Vishwakarma (WorkOnward) & Dr. Eric Wasiolek

Presented at the Frontline AI Track.

IEEE New Era AI 2025
Digital TwinStorage DevicesAI Architecture

Design Specification and AI-Driven Digital Twin Architecture for Storage Devices

Rahul Vishwakarma (WorkOnward)

Presented as a solo online talk at the Systems & Robotics Joint Symposium.

IEEE New Era AI 2025
AI SafetyData AugmentationLLMs

HarmAug: Effective Data Augmentation for Knowledge Distillation of Safety Guard Models

Seanie Lee*, Haebin Seong*, Dong Bok Lee, Minki Kang, Xiaoyin Chen, Dominik Wagner, Yoshua Bengio, Juho Lee, Sung Ju Hwang

Proposed effective data augmentation strategies for training robust safety guard models in LLMs.

ICLR 2025
EthicsVulnerabilitiesSecurity

BiasJailbreak: Analyzing Ethical Biases and Jailbreak Vulnerabilities in Large Language Models

Isack Lee*, Haebin Seong*

A comprehensive analysis of how ethical biases in LLMs correlate with their susceptibility to jailbreak attacks.

arXiv

Collaborate With Us

Interested in contributing to our research? We welcome collaborations from researchers, practitioners, and organizations who share our vision for responsible technology.