I am an AI and cybersecurity researcher based in
Chicago, Illinois, specializing in large language
model (LLM) security systems, blockchain technology,
and real-time NLP applications.

I hold a Master of Science in Computer Science from
Monroe University, New Rochelle, New York (GPA 3.8/4.0,
December 2025), with five IBM Cybersecurity
certifications and hands-on expertise across the full
AI and blockchain technology stack.

━━━━━━━━━━━━━━━━━━━━━━━━

PUBLISHED RESEARCH

CyberTrust AI: An LLM-Based Framework for Automated
Smart Contract Vulnerability Detection, Classification,
and Remediation

My flagship research project — a production-deployed
security analysis platform that applies Anthropic's
Claude Sonnet to perform contextual, cross-function
vulnerability detection on Solidity smart contracts.
In a preliminary evaluation on 35 annotated contracts
from the SWC Registry, the system achieved 100%
detection rate versus 83% for rule-based tools like
Slither and Mythril. The advantage is concentrated
in indirect reentrancy patterns and access control
logic errors — precisely the vulnerability classes
behind the most costly DeFi exploits including the
DAO hack, Euler Finance ($197M), and Nomad Bridge
($190M).

The system generates AI-powered fix suggestions,
severity-classified findings, and on-chain NFT trust
certificates for verified contracts.

Preprint: doi.org/10.5281/zenodo.19381333
Submitted to IEEE ICTAI 2026
Live platform: cybersheild-sooty.vercel.app

━━━━━━━━━━━━━━━━━━━━━━━━

PitchSense: A Real-Time NLP-Driven AI Coaching
System for Sales Call Optimization

A production-deployed SaaS platform delivering
real-time AI coaching to sales representatives
during live calls. The system captures speech via
Web Speech API, classifies prospect utterances as
objections, buying signals, or neutral using Claude
Sonnet semantic reasoning, and renders coaching
responses in under 2 seconds.

Evaluation across 12 participants demonstrated:
- 87.3% classification accuracy vs 61.3% keyword baseline
- Mean end-to-end latency of 1.73 seconds
- System Usability Scale score of 81.4
- +47% improvement in objection handling ease
- +31% improvement in call confidence

Preprint: doi.org/10.5281/zenodo.19340475
Submitted to IEEE ICTAI 2026
Live platform: pitchsense-v2.vercel.app

━━━━━━━━━━━━━━━━━━━━━━━━

TECHNICAL EXPERTISE

Security: Smart Contract Auditing · LLM Security
Analysis · Network Intrusion Detection ·
Vulnerability Assessment · Cryptography

AI/ML: Large Language Models · NLP · TensorFlow ·
PyTorch · HuggingFace Transformers · RAG Architecture

Blockchain: Solidity · Ethereum · Web3.js ·
Hyperledger Fabric · DApp Development · NFTs

Engineering: Next.js · React · Python · Node.js ·
PostgreSQL · Supabase · Vercel Edge Runtime

━━━━━━━━━━━━━━━━━━━━━━━━

RESEARCH PROFILES

ORCID: orcid.org/0000-0009-8415-0711
Google Scholar: scholar.google.com
ResearchGate: researchgate.net/profile/Yash-Mandaviya
GitHub: github.com/grootechnoyp
Website: yashmandaviya.info

━━━━━━━━━━━━━━━━━━━━━━━━

Open to research collaborations and discussions
about LLM security, blockchain auditing, and
AI-driven cybersecurity systems.