AGI. Cybersecurity.
Cryptography.
One architecture.
I design at the seam — where autonomous agents, institutional security, and cryptographic trust all have to hold the same line. The architecture is the discipline.
Who I am.
A practitioner before a portfolio. I build the systems I write about — at Visa, in the Champlain capstone lab, and on the eMerge Americas mainstage.
I am a cybersecurity architect who shipped an enterprise LLM to production before most of the industry agreed it could be done.
— Practitioner. Architect. Not a candidate looking for a chance.
I came up through Visa — Client Care senior support, then GenAI Ambassador+, then architected the RAG platform and led training and evaluation for Visa's first in-house LLM. RAG over 1,200+ knowledge sources, trained on real, non-synthetic, compliance-controlled sensitive data using reinforcement learning and continuous retesting. 37 percent to over 90 percent case-summarization accuracy in 30 days, with full OFAC compliance automation across one of the world's largest payment networks.
I won the 2024 Client Care Excellence Award not for the model — for the humans adopting it. I taught Risk, Compliance, CCPA, GDPR, OFAC, FinTech, tokenization, and global financial procedure end-to-end. I led penetration tests, malware analysis, and zero-trust cryptographic enforcement on internal systems. I shipped a Python suggestive-reasoning module from the 2025 internal hackathon straight into the Visa 365 Copilot pipeline — one of the rare hackathon entries that actually went to production.
I am finishing a B.S. in Cybersecurity at Champlain College on a 4.0 GPA, Class President, Honors League — with a 4.0 in Ethical Hacking completed June 2026. My capstone — the Sentient Sync Engine — was presented at eMerge Americas, April 23, 2026. Next: a Master of Business Science to bridge cybersecurity architecture with enterprise and financial strategy.
Three areas of active study. Each one extends the architecture — it does not dilute it.
What I do.
Three disciplines, one architecture. The seam between them is where my work lives — and it is the seam most teams cannot staff.
AGI Architecture
Production-grade orchestration for autonomous agents. LangGraph-native pipelines with collusion detection and cryptographic state commits. Sentient Sync mitigates six of the OWASP LLM Top 10 by design: LLM01 Prompt Injection, LLM02 Insecure Output, LLM04 Model DoS, LLM06 Sensitive Disclosure, LLM08 Excessive Agency, LLM09 Overreliance. Training-data, supply-chain, plugin, and model-theft risks are out of scope and addressed by complementary controls.
Cybersecurity
Risk, Compliance, CCPA, GDPR, OFAC, FinTech. Penetration testing, malware analysis, network reconnaissance, vulnerability assessment, zero-trust enforcement. I trained the humans who run it, then I shipped the systems they run.
Cryptography
HMAC-SHA256-authenticated LLM inference, Ed25519-signed, append-only tamper-evident audit trails across the full LangGraph state graph, hybrid post-quantum (Kyber-768 / ML-KEM-768) key establishment at service boundaries. Mapped to NIST CSF 2.0, OWASP LLM Top 10, and IEEE P2842.
Every transition signed. Every state commit authenticated.
A position on privacy.
I do not maintain a LinkedIn. I do not use Indeed. I do not list my profile on the platforms that monetize the surveillance of the people they claim to serve. Controlling where my information lives — and who can reach me — is not a quirk. For a cybersecurity professional, it is the definition of privacy and security, applied to myself.
I am a cybersecurity professional. My privacy, my security, and my anonymity are not optional. They are the discipline.
Platforms that aggregate professional identity for resale are incompatible with the trust model I architect inside enterprise systems. A regulator fined LinkedIn €310 million for processing its own members' data without a valid lawful basis. I will not ask employers or clients to take cryptographic signing seriously while I hand the same brokers my metadata for free.
The exposure is not hypothetical — I have lived it. List yourself once and the phone does not stop ringing, the inbox fills with recruiters who pass your personal details around without ever asking, and your name becomes a lead someone else trades. A single scraped dataset already put 700 million LinkedIn profiles — roughly 92% of users — on a hacker forum. The platforms now run on synthetic identity: ~200 million fake accounts removed in one year, and 1 in 4 candidate profiles projected to be fake by 2028. When you cannot tell which recruiter is real, the channel itself is the vulnerability — and you no longer control who reaches you.
And the network effect is a myth. LinkedIn exited China entirely — its third-largest market — and across the German-speaking world it still trails regional networks. "Being everywhere" is not reach; it is attack surface. I treat my own surface the way I treat the ones I'm hired to defend: minimize exposure, authenticate the channel, sign the message.
Then there is the bill. These platforms charge you to participate in the noise they created — roughly $30–40 a month for a job-seeker subscription, and $10,000+ per recruiter seat each year on the hiring side, all of it underwriting an inbox where, by recruiters' own accounts, the majority of outreach is spam or off-target. Meanwhile 75% of applications get no response at all, and one in five listings is a ghost job that was never real. I will not pay a subscription for the privilege of being data-mined, mass-messaged, and auto-filtered. Privacy and identity are not a premium tier — they are the baseline.
So my channel is direct, and it is mine — no subscription, no broker, no middleman taking a cut of my own name. If you want to arrange a meeting, an interview, or a visit — write to me. I read every message. I respond to every one that is real.
eMerge Americas 2026.
Miami · April 22–24, 2026. Three days inside the largest tech conference in the Western Hemisphere — what I saw, who I met, and what it changed about the architecture.
eMerge 2026 was the year the industry stopped asking whether AI was real and started asking who was going to secure it. The room was different. The people in it were different. And the language — "production-grade," "execution-ready," "non-optional security layers" — was the same language I had been writing into the Sentient Sync Engine for eighteen months.
The most technically immersive day was Wednesday, April 22 — the private AI + Quantum Computing Masterclass, eMerge's first-ever dedicated pre-conference day. Practitioner-led sessions, hands-on, directly aligned with my Sentient Sync and post-quantum work. The faculty was a who's-who of applied AI: Emre Okcular (OpenAI), Bart Czernicki (Microsoft), Rohit Patel (Meta), Joshua Lochner (Hugging Face), and Thiago E. Ferreira (Elevate AI). The takeaway was unambiguous: the industry has shifted from AI experimentation to AI execution. Frameworks have to be production-ready, not prototype-grade.
"The industry has shifted from AI experimentation to AI execution — frameworks have to be production-ready, not prototype-grade." — the unambiguous takeaway from the AI + Quantum Computing Masterclass.
Thursday, April 23 was the main stage — Brad Garlinghouse (Ripple), Mariana Atencio, Johan Gerber (Mastercard), Colin Mahony (Recorded Future), Saswat Panigrahi (Waymo). Waymo confirmed Miami as their 11th full-service autonomous market — 500,000 weekly autonomous trips, 10× growth in two years. The city I work in is now a live testing ground for autonomous systems. The architecture I am building is for exactly this environment.
On the AI + Deep Tech stage I met **Peter Yared** of AgentCloak — agent identity, data residency, AI security. His work maps directly onto the auditor.py logic and the agent collusion detection framework. NVIDIA's Vishal Ganeriwala reinforced that GPU-accelerated inference is now the backbone of any production-grade agent deployment. Across the conference, the throughline was clear: institutional funding and enterprise demand are converging on exactly the AI-security and post-quantum work the Sentient Sync Engine is built for.
Friday, April 24 — the Florida Innovation Ecosystem Breakfast, 300+ exhibitors, six stages running in parallel. This was the day I presented the Sentient Sync Engine live. The feedback I returned with is why this site looks the way it does now.
Three strategic conclusions I left Miami with:
- —AI has moved from pilot to production. My architecture must be execution-ready, not demo-grade. The site you are reading reflects that.
- —AgentCloak is the closest competitive intelligence I have, and a potential collaboration. The space is real and someone else is already in it.
- —The funding and enterprise demand are converging on AI + cyber + quantum work. The Sentient Sync Engine is built for exactly that environment.
Live threat intel.
Direct feed from CISA's Known Exploited Vulnerabilities catalog — refreshed hourly, mapped to OWASP Top 10 2021 via CWE. The dashboard isn't a widget. It is a working surface I read every morning.
Recent entries
Reading the catalog…
Selected work.
A short list, by design. I do not publish every assignment. These are the engagements where the architecture, the cryptography, and the AI all had to hold the line together.
Visa's first in-house LLM
Visa 365 Copilot — suggestive reasoning module
Enterprise Incident Response Framework
The Sentient Sync
Engine.
My Champlain capstone and the project I presented at eMerge Americas 2026. A production-grade AGI security framework that detects multi-agent collusion in real time — and signs every line of evidence on its way out.
Real-time agent collusion detection.
Behavioral fingerprinting across multi-agent communication channels. Hybrid post-quantum (Kyber-768 / ML-KEM-768) key establishment at every service boundary. Ed25519-signed, append-only tamper-evident audit trails across the full LangGraph state graph. Mapped natively to NIST CSF 2.0, OWASP LLM Top 10, and IEEE P2842.
Secure Ingestion → Agent Orchestration → Cryptographic Signing → NIST-Aligned Governance → Gemini Threat Analysis. Every transition signed. Every state commit witnessed.
Cryptographically signs every flagged event before state commit. Tamper-evident inferencing on the same pipeline as the audit layer.
Structured JSON threat analysis over flagged logs. HMAC-SHA256-authenticated before commit. Tamper-evident LLM inference.
LLM01 Prompt Injection · LLM02 Insecure Output · LLM04 Model DoS · LLM06 Sensitive Disclosure · LLM08 Excessive Agency · LLM09 Overreliance.
Arrange a meeting.
Email is the channel of record. No platforms, no aggregators, no auto-rejected pipelines. Write — I read every message.
Open to associate and junior security / AI roles while completing my degree.
For meetings, interviews, project inquiries, and visits. Responses typically within 24 hours, Eastern Time.
Routed through this domain. Use for press, conference programming, or anything that should live on the public record.
EU & US work authorized · Open to relocation across the European Union and the United States.