Privacy-Preserving AI
Designing machine learning architectures prioritizing local inference, hardware security boundaries, and explainable risk models without sacrificing capability.
Researcher · Software Engineer · Educator
Avijit Roy
I build and study secure, privacy-aware AI systems at the intersection of cybersecurity, digital forensics, data infrastructure, and human-centered computing. As an educator, researcher, and data analyst, I focus on translating complex technical ideas into practical systems, applied research, and real-world learning experiences.
In the News
Institutional Feature · John Jay College
Avijit Roy ’21, ’25, ’26: Computer Scientist Protecting People in the Age of AI
John Jay College featured my journey as a computer scientist, AI and cybersecurity researcher, educator, and CUNY data professional working to protect people in the age of AI.
Read the FeatureProfessional Thesis
I am originally from Bangladesh and now based in New York City. My work brings together cybersecurity, digital forensics, artificial intelligence, and applied data systems. Across these areas, I am most interested in building technology that solves real problems without losing sight of privacy, security, and the people using it.
I currently serve as a full-time Substitute Lecturer in Computer Science at John Jay College of Criminal Justice (CUNY) and as an Adjunct Assistant Professor at St. John's University, where I teach a graduate course in cyber forensics and malware analysis. I also work as a Data Analyst at the CUNY Central Office. In the classroom, I focus on helping students understand how systems work; at CUNY, I build reliable reconciliation and validation workflows for complex, multi-campus data.
My research focuses on the ways AI and security systems can fail under adversarial, linguistic, and privacy-related constraints, and how those systems can be made more robust and trustworthy. My master's thesis examined the adversarial robustness of perceptual hashing systems. I am also continuing research in machine learning and low-resource language infrastructure, supported by computational allocations from NSF ACCESS, on which I serve as Principal Investigator.
Outside of work, I am an avid traveler and photographer. Travel has given me a broader perspective on people, environments, and the different ways systems are experienced in practice. That perspective continues to influence how I think about technology, research, and the kinds of problems worth solving.
Education
-
MS — Digital Forensics & Cybersecurity (D4CS), John Jay
College
(CUNY), 2026
Claude Hawley Medal · Awarded for the best master's thesis and/or scholastic distinction in the graduate programGraduate Achievement Award · Commencement award recognizing perseverance, leadership, and impact within the graduate communityStudent Council Scholarship · Recognized for academic dedication, leadership, resilience, and community impact
- Advanced Certificate — Applied Digital Forensic Science (CAD4SCI), John Jay College (CUNY), 2025
-
BS — Computer Science & Information Security, John Jay
College (CUNY), 2021
Leonard E. Reisman Medal · Outstanding Senior for Distinguished Scholarship & Service
Teaching & Mentorship
I teach applied computer science with a focus on clear algorithmic thinking and secure systems. Mentoring students to build robust software is a cornerstone of my professional practice.
Current Teaching
John Jay College · CUNY
Fall 2026 Object-Oriented Programming · Database & Data Mining · Computer Algorithms
St. John's University
Fall 2026 · Graduate Cyber Forensics & Malware Analysis
Books
Long-form guides for general readers and students alike.
Kindle Edition
Cybersecurity Fundamentals: A Practical Guide for Everyone
A jargon-free guide to online safety, scams, and identity protection for everyday readers — available now on Amazon Kindle.
Learn More
Free ebook
Saving Your Work Like a Pro: Git & GitHub for C++ Students
A print-ready, 17-chapter guide to version control for first-semester C++ students.
Read the Ebook
Focus Areas
Where teaching, research, and engineering converge into practice.
Cyber & Forensics
Developing secure systems, anti-phishing safeguards, and immutable evidence validation protocols for investigative analysis.
Applied Data Systems
Enabling institutional scalability through .NET backends, optimized SQL data warehouses, and rigorous CI/CD cloud integrations.
Technical Mentorship
Mentoring the next generation of engineers to master programmatic rigor, algorithmic thinking, and ethical software architecture.
Academic Research
Advancing the intersection of cybersecurity, applied machine learning, and public-interest intelligence systems.
Published by Wiley · Applied Research · 2026
AI-Driven and Cyber-Secure Embedded Inverter Control for Energy Optimization in Electric Vehicles
A deep-reinforcement-learning inverter control framework with an embedded intrusion detection system, validated on WLTP and UDDS driving cycles and published in a Scopus-indexed, Q2 Wiley journal.
View Published Paper
Published by Springer · Best Paper Award · ICICC 2026
Privacy-Preserving MIS Architectures
Integrating blockchain validation, federated learning, and zero-trust models to enable secure distributed data collaboration while minimizing centralized exposure.
View Published Paper
FLAIRS-39 Oral Presentation
Propasafe-Hybrid
A hybrid propaganda detection system combining local BERT-based sentence classification, selective LLM explainability, and cost-aware filtering for practical media analysis workflows.
View Presentation
Patent Pending
Client-Side Phishing Risk Assessment
A localized, staged phishing detection architecture utilizing onboard ONNX transformer inference to evaluate risks entirely within the email client runtime.
View ProtocolEngineering Portfolio
Scalable toolchains and rigorous system deployments.
Legacy Intelligence Platform
Capstone Legacy System Modernization. Modernized a nonfunctional .NET 5 and React investigation platform by migrating its backend to .NET 9, rebuilding broken components, restoring real-time geospatial workflows, and deploying the enhanced system to Microsoft Azure.
View DetailsDjango Auth
Strict Session Enforcement. An asynchronous authentication system relying on deterministic token invalidation and GeoIP logging to prevent concurrent testing access.
View SourceArticleExtractor
NLP Data Cleanse Protocol. A robust extraction engine that aggressively strips HTML noise to yield perfectly normalized datasets utilized for sentiment and linguistic analyses.
View DetailsHashNow
Zero-Trust Verifier. A local WASM-powered browser tool calculating MD5, SHA-1, and SHA-256 signatures client-side, guaranteeing absolute document confidentiality.
Test ToolContact
Get in Touch
I enjoy connecting with technologists, educators, and storytellers. Whether discussing architectural design patterns or sharing global travel experiences, I'd be glad to hear from you.