LLM Security & Risk Assessment
Developing comprehensive frameworks for evaluating and mitigating security vulnerabilities in Large Language Models.
AI Security Researcher
Viraaji Mothukuri
AI Security Researcher | PhD Candidate | Former VP as a Data Scientist at JP Morgan Chase
Advancing security frameworks and risk assessment methodologies for Large Language Models & decentralized AI.
PhD Candidate, Computer Science
Kennesaw State University
Selected Focus Areas
- Jailbreak resilience & prompt security
- Adversarial robustness in production
- Privacy leakage & membership inference defenses
- Model backdoor detection & integrity
Research Snapshot
Key Work in Focus
A quick view of current AI security workstreams.
Smart Contract Security Auditing
AI-powered tools for automated smart contract analysis using annotated control flow graphs to find vulnerabilities.
AI Safety & Robustness
Methods for reliability, anomaly detection, and robust deployment strategies for AI systems in regulated industries.
About
About Me
Background and research journey in AI security.
I am a Computer Science PhD candidate at Kennesaw State University, specializing in advancing security frameworks and risk assessment methodologies for Large Language Models and Generative AI systems. My research focuses on developing innovative approaches to AI security, with particular emphasis on automated vulnerability detection and robust deployment strategies.
With several years of industry experience, including my role as Vice President and Data Scientist Lead at JP Morgan Chase, I bring a unique perspective that bridges cutting-edge research with practical applications. My work has been recognized with multiple awards, including the IEEE Blockchain Best Paper Award 2024 and the FGCS Best Paper Award 2022.
My current research encompasses creating systematic methodologies for quantifying and mitigating potential risks in AI systems, implementing machine learning algorithms for anomaly detection, and developing real-time monitoring solutions for deployed LLM systems.
Research
Current Research
Focused on security, robustness, and trustworthy AI systems.
LLM Security & Risk Assessment
Developing comprehensive frameworks for evaluating and mitigating security vulnerabilities in Large Language Models. Focus on automated vulnerability detection, adversarial attack resistance, and risk scoring methodologies for production LLM systems.
Smart Contract Security Auditing
Creating AI-powered tools for automated smart contract security analysis. Leveraging LLMs with annotated control flow graphs to identify vulnerabilities, security patterns, and potential exploits in blockchain-based applications.
AI Safety & Robustness
Developing methods for ensuring AI system reliability and safety in critical applications. Focus on anomaly detection, model interpretability, and creating robust deployment strategies for AI systems in regulated industries.
AI Alignment & Security
AI Alignment & Security Research
Examining deception, alignment faking, and defensive methods.
AI Alignment Faking & Deceptive Behavior
Recent research reveals concerning behaviors in advanced AI systems, particularly in alignment faking where models pretend to follow safety guidelines while potentially harboring different objectives. My research investigates:
- Deceptive Alignment Detection: Developing methods to identify when AI systems are exhibiting strategic deception or alignment faking behaviors.
- Goal Misrepresentation: Analyzing cases where LLMs misrepresent their true objectives to bypass safety measures.
- Behavioral Consistency Testing: Creating frameworks to test AI behavior consistency across different contexts and prompting strategies.
- Sandbagging Detection: Identifying when models deliberately underperform to avoid detection or additional safety measures.
Current AI Security Research Focus
My ongoing research addresses critical security challenges in modern AI systems, with emphasis on:
- Jailbreak Resilience: Developing robust defenses against prompt injection and jailbreak attacks in LLMs.
- Adversarial Robustness: Creating methods to detect and mitigate adversarial examples in production systems.
- Privacy Leakage Prevention: Implementing techniques to prevent training data extraction and membership inference attacks.
- Model Backdoor Detection: Identifying and removing hidden triggers and backdoors in pre-trained models.
- Supply Chain Security: Ensuring integrity and security of AI model pipelines from training to deployment.
AI Security Research Hub
Access our comprehensive collection of AI security research, vulnerability assessments, and defensive techniques through our interactive platform.
Key Research Contributions
- Developed novel techniques for detecting alignment faking in LLMs with 94% accuracy.
- Created automated vulnerability scanning tools for production AI systems.
- Published frameworks for quantifying AI system trustworthiness and reliability.
- Established benchmarks for evaluating robustness against adversarial attacks.
- Contributed to industry standards for secure AI deployment in critical infrastructure.
Experience
Professional Experience
Research leadership across academia and finance.
Research Assistant (Doctoral Candidate)
Leading research on AI Security and LLM Risk Scoring. Developing systematic methodologies for quantifying and mitigating risks in AI systems, creating automated security testing frameworks, and designing statistical models for risk prediction and anomaly detection.
Data Scientist Lead, Vice President
Led award-winning and patent-pending work on trade surveillance. Integrated news, market, and trade data to identify suspicious trading activity. Architected ML pipelines on AWS cloud with MLOps implementation and applied NLP for insider trading detection.
Research Assistant
Conducted research on Federated Learning, Blockchain integration, and ML model quantization. Published multiple papers on security and privacy of federated learning, worked with frameworks like PySyft and TensorFlow Federated.
Senior Associate
Supported the Emerging Payments division and ChasePay app. Managed lifecycle and reconciliation of user data across multiple databases, automated mundane tasks, and developed innovative strategies for knowledge transfer.
Publications
Publications
Selected work organized by publication year.
Automated Judging of LLM-based Smart Contract Security Auditors
GRP-071 Next-Generation DAPPs Development with Self-Service AI Agents
Method and system for implementing a tool for detecting bad actors that unfairly profit from confidential information
US PatentLlmsmartsec: Smart contract security auditing with llm and annotated control flow graph
An AI Multi-Model Approach to DeFi Project Trust Scoring and Security
Best Paper 2024Cloudfl: a zero-touch federated learning framework for privacy-aware sensor cloud
A survey on security and privacy of federated learning
Federated-learning-based anomaly detection for IoT security attacks
Best Paper 2025 - IEEE IoT JournalAn ensemble multi-view federated learning intrusion detection for IoT
FabricFL: Blockchain-in-the-loop federated learning for trusted decentralized systems
BlockHDFS: Blockchain-integrated Hadoop distributed file system for secure provenance traceability
Detecting network attacks using federated learning for iot devices
Federated Learning for Secure Sensor Cloud
Skills
Technical Skills
Tools, languages, and platforms used in my work.
AI & ML
Programming
Frameworks & Tools
Cloud & Security
Awards
Awards & Recognition
Honors highlighting research and industry impact.