Executive Summary
Cybersecurity undergraduate at Old Dominion University (3.9 GPA) with CompTIA Security+, Network+, and A+ certifications. AI security researcher with two arXiv preprints on synthetic fraud detection (COVA framework, under review at IEEE DASC and IEEE BigData 2026) and a live dataset platform (ScamLingua.org). Background includes over a decade in high-risk, regulated financial environments (FNMA, FHLMC, FHA, VA), applying the same rigor to incident response, system hardening, and AI governance. Proven record of managing 3× standard workloads while sustaining 95%+ accuracy.
🔬 Featured Research
Mentored by Dr. Ayan Roy (Christopher Newport University) • CCI Undergraduate Research Program
COVA Framework
Under ReviewMulti-agent LLM framework for generating labeled synthetic scam conversations targeting elder populations. Established baseline classification using XGBoost + TF-IDF achieving 72.5% accuracy across 8 elder-fraud categories.
▶ Technical Details
🛠️ Tools & Frameworks
🔬 Methodology Highlights
- Two independent LLM agents (attacker + victim) exchange turns iteratively, producing natural resistance and escalation dynamics across 8 elder-fraud categories and 16 prompt templates.
- Parameterized victim personas (age 65–85) with trust, scam-awareness, and tech-savviness traits; attacker knowledge tiers (cold-call 50% / partial 30% / full 20%).
- Initially prototyped with commercial LLM APIs (including Anthropic Claude); transitioned to local Qwen 2.5 14B inference via Ollama on NVIDIA RTX 4080 Super (16 GB VRAM) after safety guardrails blocked scam role-play scenarios — enabling unrestricted generation, zero API cost, and full data privacy.
- Built a negation-aware v2 label-audit pipeline after discovering a 49.8% initial mismatch, relabeling 1,594 conversations for outcome-label integrity.
📊 Notable Results
- Benchmarked 8 models (6 classical ML + 2 transformers); XGBoost + TF-IDF led at 72.5% accuracy / 0.691 macro-F1.
- TF-IDF lexical features beat 28 hand-engineered features by 16 points — quantifying how much discriminative signal lives in the raw text.
- Isolated input truncation and limited data scale as the transformer bottleneck — a hypothesis directly tested (and confirmed) in COVA-X.
🏗️ Pipeline Architecture
End-to-end pipeline: prompts → local LLM → multi-agent dialogue → quality checks → labeled dataset
🖼️ Pipeline in Action
Live pipeline screenshots: generation → sample dialogue
👤 My Contributions
Lead author. I designed and implemented the multi-agent generation pipeline that produces realistic attacker-victim dialogue exchanges. I engineered the 28-feature extraction system and TF-IDF representation, built the v2 label-audit tooling that corrected 1,594 mislabeled conversations, and trained and evaluated all eight baseline classifiers.
COVA-X Benchmark
Under ReviewExpanded benchmark with 10,985 synthetic conversations. Fine-tuned Longformer achieving 79.71% accuracy and 0.779 macro-F1, advancing the state of elder-fraud detection research.
▶ Technical Details
🛠️ Tools & Frameworks
🔬 Methodology Highlights
- Scaled the corpus 3.4× to 10,985 conversations (88,852 turns) with five distinct victim/attacker profiles per scam type for behavioral diversity.
- Engineered a three-role generation architecture for virtual-kidnapping scenarios, cutting artifact flag rates from 67.1% to 46.5%.
- Built a multi-stage quality lifecycle — contamination scanning, stage-direction stripping, and automated relabeling — driving a 12.7× label-consistency gain (49.8% → 3.9% correction rate).
- Characterized Qwen 2.5 14B capability limits under sustained emotional pressure, including a salience-bias effect where enumerating forbidden patterns increased their production.
📊 Notable Results
- Longformer overtook XGBoost on every metric (79.71% accuracy / 0.779 macro-F1 vs. 78.43% / 0.756), directly confirming the first paper's data-scale hypothesis.
- Transformers gained the most from scale (Longformer +11.2 macro-F1 points vs. XGBoost +6.5), with the biggest edge on the minority "complied" class.
- Key challenge identified: distinguishing partial compliance from rejection — victims who engage extensively before refusing present similar lexical patterns to those who partially comply, making this the dominant error boundary across all architectures.
- A pre/post-cleanup sensitivity analysis showed all three architectures improved — evidence the pipeline recovered genuine label-relevant signal, not architecture-specific noise.
🖼️ Scaling to 10K+
Scaling & QA screenshots: quality scan → final dataset
👤 My Contributions
Lead author. I scaled the generation pipeline across dual-GPU workstations (RTX 4080 Super + RTX 5060 Ti), designed the three-role virtual-kidnapping architecture that reduced artifact rates by 20 percentage points, and built the complete multi-stage quality-lifecycle tooling — including contamination scanning, stage-direction stripping, and automated relabeling. I then retrained and evaluated all classifiers to validate our data-scale hypothesis.
ScamLingua Platform
LiveSelf-built distribution platform for COVA research datasets and benchmarks. Designed with trust-focused web engineering principles to support reproducible AI safety research.
🎬 COVA Framework Demo
🎥
Demo Video Coming Soon
Multi-agent conversation generation in action
Watch the COVA framework generate realistic attacker-victim conversations in real-time.
🗓️ Research Timeline
-
COVA Framework & Dataset
First public multi-turn smishing dataset — 3,201 conversations, 8-model baseline (XGBoost + TF-IDF, 72.5%). arXiv:2604.11752
-
COVA-X Expanded Benchmark
Scaled to 10,985 conversations; Longformer surpasses XGBoost at 79.71% accuracy, confirming the data-scale hypothesis. arXiv:2606.06879
-
ScamLingua Platform
Self-built distribution platform bringing the COVA research datasets and benchmarks to the broader AI-safety community.
-
COVA-B Benign Dataset
In ProgressCompanion dataset of legitimate conversations for precision benchmarking.
-
Multi-Model Comparison Study
PlanningCross-model evaluation of LLM capability ceilings in synthetic dialogue generation.
📊 Technical Skills
-
Certifications:
CompTIA Security+, Network+, A+ -
AI/ML Research:
Multi-Agent LLMs, Synthetic Data Generation, Transformer Fine-Tuning, Prompt Engineering, Dataset Engineering, Ollama/Qwen Local Inference -
ML Frameworks:
Hugging Face Transformers, PyTorch, scikit-learn, XGBoost, Longformer/DistilBERT, TF-IDF -
Cybersecurity:
Incident Response, System Hardening, Network Defense, AI Governance, Vulnerability Assessment -
Tools:
Nmap, Wireshark, pfSense, tcpdump, Snort, iptables, SSH -
Development:
Python, Trust-Focused Web Engineering, HTML/CSS/JS, SQL, GitHub -
Platforms:
Kali Linux, Ubuntu, Debian, BSD/pfSense, Windows 10/11, Windows Server, home lab environment
📊 Skill Distribution
🧠 Education
Old Dominion University
B.S. Cybersecurity | GPA: 3.9
CCI Undergraduate Research Program
Jax Code Academy
Web Development | 2023 Graduate
Roanoke College
B.A., Criminal Justice & Philosophy
💻 Featured Project
Secure File Sharing System
Python TCP Sockets • Milestone Project
Problem Addressed
Organizations routinely transfer sensitive data over insecure channels, creating exposure to interception and tampering. This project explores secure design principles for client–server exchange.
System Design
- Custom TCP client–server architecture in Python
- User authentication & directory isolation (DAC)
Security Controls
- Encrypted communication channel
- Data‑integrity validation during transfer
🔧 Personal Projects
LockBadges
Windows Desktop Utility • Shipped & Documented
Movable, translucent on-screen indicators for Caps Lock, Num Lock, Scroll Lock, and Mute — built for keyboards that ship without status LEDs. Each badge is independently positioned, styled, and configured through a full settings GUI, renders click-through, and hides automatically during fullscreen apps.
Privacy by construction: lock state is read from Windows toggle bits rather than a keyboard hook, so keystrokes are never intercepted — with no network access, no telemetry, and no logging.
🛡️ Labs & Applied Coursework
Incident Response & Forensics (NIST SP 800‑61)
Executed full IR lifecycle after simulated brute‑force: recon detection with Nmap, log analysis, volatile data collection, and GPO remediation.
Operating System Hardening & Policy Enforcement
Strengthened Windows/Linux security baselines via Group Policy Objects, authentication auditing, and attack surface reduction using iptables.
Network Defense & Firewall Security (pfSense)
Applied defense‑in‑depth via firewall rule design, disabling insecure protocols, and configuring SSH for encrypted administrative access.
Intrusion Detection & Packet Analysis (Snort / Wireshark)
Deployed Snort IDS and analyzed PCAP traffic using tcpdump and Wireshark to identify anomalous behavior.
🧩 Professional Experience
Cardinal Housing, LLC | Property Management & Estate Fiduciary
Jacksonville, FL | 2012 – 2024
Self-employed for nearly 13 years, running a three-property rental portfolio while serving as fiduciary for two private estates in probate. The day-to-day work was recordkeeping, regulatory compliance, vendor oversight, and being the only person on call when something failed.
- Records & Data Integrity: Maintained financial records, lease agreements, and legal documentation to statutory standards, reconciling rent collections against landlord and resident statutes.
- Confidentiality of Sensitive Data: Safeguarded personally identifiable financial and legal information for tenants and estate beneficiaries across a 12-year retention period.
- Regulatory & Legal Compliance: Coordinated with legal counsel and financial institutions to meet probate law, tax, and contractual obligations, including asset distribution and liquidation for two estates.
- Incident Response & Availability: Held 24-hour on-call responsibility for critical building and tenant failures, restoring habitability and service under time pressure.
- Vendor Management: Selected, contracted, and supervised trades against fixed budgets, processing change requests and coverage research on repair issues.
- Risk & Financial Analysis: Identified and negotiated acquisitions using national market research, evaluating market cycles, cost projections, and budget exposure before committing capital.
BlueHub Capital (SUN) | Foreclosure Relief Underwriter
Boston, MA | 2021 – 2022
Underwrote foreclosure-relief mortgages for distressed homeowners through pandemic-era market disruption, in a lending environment subject to bank examination.
- Compliance & Audit Support: Assisted in compliance matters including bank exams and audits, protecting the financial interests of the company, its investors, and clients.
- System Configuration & Process Change: Contributed to the revision of Encompass, the loan origination platform, to streamline the application process for staff and applicants alike.
- Control & Procedure Design: Rewrote underwriting guidelines and loan-processor checklists to meet updated regulatory guidance and shifting economic conditions during COVID-19.
- Encompass platform work: Contributed to revisions of Encompass, the loan origination platform, streamlining the application process for staff and applicants.
- Sensitive Document Handling: Coordinated appraisals, title, tax transcripts, and credit supplements with third-party vendors, negotiations, and closing staff.
- Training & Enablement: Trained and mentored newly hired personnel on completing complex applications and communicating with applicants.
EverBank & TIAA Bank | Loss Mitigation Underwriter
Jacksonville, FL | 2017 – 2021
Assessed loss-mitigation documentation in a high-stakes servicing environment operating under federal consent decree, working daily across a stack of government and investor underwriting platforms.
- Regulatory Compliance under Consent Order: Sustained a 97% average on Quality Right Party Contact Monitoring Reports, the control measuring adherence to Office of the Comptroller of the Currency (OCC) Consent Orders.
- Regulated Platform Operation: Worked daily in EVARS (HUD/FHA), VALERI (VA), FNMA HSSN, and FHLMC Workout Prospector underwriting platforms, alongside CAIVRS, Black-Knight/MSP, PACER, SCRA, and LexisNexis.
- Risk Assessment & Discrepancy Detection: Analyzed loss-mitigation documentation to assess risk and identify requirement discrepancies, generating borrower notices to resolve deficiencies before foreclosure.
- Verification & Validation: Validated court-ordered requests, obtained bankruptcy approvals, and evaluated collection, loss mitigation, and foreclosure activity against investor approval matrices and state and federal guidelines.
- Mediation & Court Proceedings: Represented the bank’s foreclosure team in mediation, on calls and in hearings alongside borrowers’ counsel and presiding judges, coordinating with courts, third-party attorneys, mediators, and agents.
- Cross-Functional Interface: Facilitated system and operational interfaces between legal and business units to support compliant liquidation and mediation.
- Throughput under Load: Managed a pipeline 3× the standard workload while sustaining a 95%+ quality rating.
EverBank & Ditech Financial | SPOC Relationship Manager
Jacksonville, FL | 2011 – 2014
Single point of accountability for borrowers and third parties across the foreclosure-prevention and loss-mitigation lifecycle.
- Ownership & Accountability: Owned every aspect of client and third-party foreclosure, foreclosure-prevention, and loss-mitigation inquiries as the designated Single Point of Contact.
- Confidentiality & Accuracy: Handled sensitive financial information and legal documentation throughout the foreclosure lifecycle without disclosure or error.
- Pipeline & Deadline Management: Managed and monitored loss-mitigation package pipelines against hard deadlines, exceeding investor perfection rate requirements by 30%.
AmeriCorps Habitat for Humanity | Construction Team Leader
Jacksonville, FL | 2009 – 2011
Led daily volunteer crews on residential builds under an AmeriCorps service contract.
- Leadership at Scale: Coordinated up to 150+ daily volunteers across framing, roofing, siding, windows, drywall, trim, tile, and finish work.
- Delivery & Quality: Oversaw building and rehabilitation of 100+ Duval County homes plus a $1.2M housing project, achieving a 98% quality rating, 97% completion time, and 98.6% volunteer satisfaction.
- Crew Collaboration: Worked alongside volunteers from every background — Fortune 500 executives, accountants, military personnel, students, trade professionals, and court-ordered individuals.
- Financial Wellness Coaching: Guided participants on goal setting, budgeting, credit rebuilding, savings planning, and asset development.