AI Engineer · Cybersecurity Innovator · Gamification Pioneer

Clint Bodungen

I build

At the intersection of AI, cybersecurity, and games — building with AI since 2013, long before the wave.

Explore My Work AI résumé

2013building with AI since

30+years in cybersecurity

2published books

The arc

Cybersecurity is where I come from — innovation is what drives me.

For 30+ years I've done the same thing on repeat: bring an emerging technology into a field before it's obvious — first gamification, then AI. Find the frontier, and make it real.

Most of those years were spent directly inside the environments people usually write about second-hand: supermajor oil and gas, electric utilities, chemical manufacturing, national labs, major tech and infrastructure, and federal agencies. Stay in one field that long and you know the people who built it. When you need to reach the right person, I can usually pick up the phone.

Co-author, Hacking Exposed: Industrial Control Systems·Repeat S4 speaker

The origin

Gamification of Cybersecurity

Co-created the world's first online, multiplayer, game-based cybersecurity simulation — making security training something you play, not sit through.

Since 2013

AI in Cybersecurity

Brought AI into security tooling and training years before the generative-AI wave — and published on generative AI for security while most of the field was still watching.

The convergence

AI + Gamification

Adversarial game AI, autonomous AI exercise facilitation, and AI-driven training — the two frontiers, fused.

Now

AI in Games & AI That Builds

AI in commercial video games — and AI systems that engineer software themselves. The domain keeps changing; the pattern doesn't.

Capabilities

An unusually broad AI toolkit

From classic and bio-inspired AI to modern LLM and agentic systems — techniques applied across games, enterprise platforms, and security research.

Agentic & LLM Systems

Multi-Agent Orchestration
Agent Harnesses
Retrieval-Augmented Generation (RAG)
Persistent AI Identity & Memory

Bio-Inspired & Classic AI

Swarm Intelligence (Ant Colony Optimization)
Evolutionary Computation & Genetic Algorithms
Neural Networks & Neuroevolution
Multi-Objective Optimization

Game AI

Utility-Based Behavior Trees
Adversarial Search (Minimax)
Pathfinding and Probability Algorithms
AI-native NPCs and Narrative

Data & Knowledge

Knowledge Graphs (i.e. Neo4j)
Vector Databases & Semantic Search
LLM Fine-Tuning

01 — Portfolio

Flagship work

Named products and research, plus a genericized enterprise engagement. Every AI claim traces to real, shipped systems.

AI Infrastructure · personal

TestFlight — Multi-Agent Engineering Framework

A substrate-neutral framework that turns a coding agent into a disciplined, multi-agent software-engineering system — ~20 specialist agents, a verification loop with adversarial QA, and a differentiated rapid-prototyping methodology. Blind-judged experiments show it lifts a cheaper model to a frontier model's quality band. It built this very site.

Multi-AgentAgentic SDLCRapid PrototypingAI Eval

AI Infrastructure · personal

MindStone — Persistent AI Identity & Memory

A platform for persistent, memory-continuous AI agents: vector-backed semantic recall with experiential-salience weighting, a no-compaction continuity model, and dream-cycle consolidation — so an agent keeps its identity across sessions and even across different underlying models.

Vector MemoryRAGLLM AgentsLanceDB / sqlite-vec

Research · personal

Project DARWIN — Bio-Inspired Cyber-AI

Open research applying bio-inspired AI to cyber risk: virtual 'ant' swarms (Ant Colony Optimization) traverse a Neo4j attack graph of MITRE ATT&CK / D3FEND / CVSS to surface the most probable, high-impact attack paths, while genetic algorithms evolve adaptive defenses.

Swarm IntelligenceGenetic AlgorithmsNeo4jATT&CK

Product · ThreatGEN

ThreatGEN® Red vs. Blue

The world's first online, multiplayer, game-based cybersecurity simulation. Its red-team/blue-team opponent runs a utility-based game AI — behavior trees driven by animation-curve utility scoring — re-engineered from Unity/C# into a TypeScript web app.

Game AIUtility Behavior TreesUnity → Web

Product · ThreatGEN

ThreatGEN AutoTableTop™

An AI-powered incident-response tabletop-exercise platform driven by a multi-agent LLM system: an AI facilitator, scenario/timeline agents, and dynamic 'inject' delivery with structured outputs and real-time orchestration.

Multi-Agent LLMStructured OutputReal-time

Product · personal

Dreamscape Legends

A commercial trading-card game with a minimax adversarial-search opponent (scalable difficulty) plus production LLM features — in-game AI narrative characters and AI-generated campaign content — and AI-assisted game design and balance analysis.

MinimaxLLM / GenAIGame Design

Product · personal

Operation Zero Hour

A full NFC conference-gamification platform: a Unity mobile app (player + sponsor lead-capture), a Firebase backend, and a React admin dashboard with conference-management features — guest-list import, badge-PDF generation, walk-up registration, and analytics.

MobileFirebaseReactFull-Stack

Enterprise · critical-infrastructure security client

AI Vulnerability-Management Platform

client

A security knowledge graph (Neo4j) fusing enterprise assets, vulnerabilities, MITRE ATT&CK / D3FEND, and CISA KEV exploit intelligence, with a multi-agent AI layer for natural-language querying and a four-layer AI asset-deduplication engine. Delivered under a fixed-price engagement.

Knowledge GraphLangGraphRAGNL→Cypher

Experience

AI résumé

Leading AI/ML engineering, pioneering AI in cybersecurity products, and self-directed AI research — a track record that predates the current AI wave by a decade.

Arcova

formerly MorganFranklin Cyber · Full-time · Remote

Director, AI/ML Engineering

Aug 2025 – Present

Lead the AI/ML engineering team building production, agentic-AI applications across cybersecurity and GRC — including an AI-powered third-party risk-management platform (automated vendor tiering, evidence mapping from trust centers, OSINT/dark-web risk monitoring), an AI change-management intake system, and LLM-powered go-to-market and sales-intelligence tools.
Architect multi-agent and LLM systems using retrieval-augmented generation (RAG), structured/constrained generation, tool-using research agents, and provider-abstracted multi-model integration (Anthropic, OpenAI, and local models).
Established an AI-efficacy testing framework — multi-layer, including LLM-as-judge evaluation — as an engineering release gate that measures whether AI features actually perform, not just whether the code runs.
Advise on the secure integration of agentic AI — human-in-the-loop controls, source provenance/confidence tracking, and deterministic fallbacks.
Drive AI-native product strategy, reframing document-heavy GRC workflows around AI ingestion/inference layers that draft analysis from evidence with per-field provenance.

Director, Cybersecurity Innovation

Aug 2024 – Aug 2025

Led an innovation engagement delivering an AI-powered vulnerability-management platform for a critical-infrastructure security client — fusing enterprise assets, NVD vulnerabilities, MITRE ATT&CK / D3FEND, and CISA KEV exploit intelligence into a Neo4j security knowledge graph queryable in natural language.
Designed a multi-agent AI layer (LangChain / LangGraph): natural-language-to-Cypher graph querying, RAG-based security chatbots, and a self-extending meta-agent that writes its own graph analytics from plain-English use cases.
Built an exploit-aware, four-layer AI asset-deduplication engine (deterministic fingerprinting → semantic-embedding similarity → graph adjacency → LLM adjudication) that cut LLM cost ~70% via cheapest-method-first tiering.

ThreatGEN

Founder / Chairman / Head of Product Innovation · Full-time

Founder / Chairman / Head of Product Innovation

Jul 2017 – Present

Founded a funded cybersecurity startup closing the skills gap through gamification and AI-driven training — built on modern game engines, simulation technology, and Generative AI/LLMs.
Co-creator of ThreatGEN® Red vs. Blue — the world's first online, multiplayer, game-based cybersecurity simulation — including its adversarial red-team/blue-team game AI: utility-based behavior trees with animation-curve utility scoring (plus an exploratory neural-network opponent), now re-engineered for the web in TypeScript.
Creator of ThreatGEN AutoTableTop™ — an AI-powered incident-response tabletop-exercise platform driven by a multi-agent LLM system (an AI facilitator, scenario/timeline agents, and dynamic inject delivery) with structured outputs and real-time orchestration.
Pioneering the applied use of Generative AI and LLMs across cybersecurity training, exercises, and real-world application — spanning game AI, autonomous exercise facilitation, and AI-assisted content generation.

Independent AI Research

Ongoing

Founder & Principal Researcher

Building with AI since 2013

Designed a persistent-identity and long-term-memory architecture for LLM agents (Layered Continuity Architecture) — vector-backed semantic recall with experiential-salience weighting and dream-cycle consolidation — enabling agents to keep identity and knowledge across sessions and across different underlying models.
Built a substrate-neutral, multi-agent software-engineering framework that orchestrates ~20 specialized AI sub-agents under a codified SDLC, with a rapid-prototyping methodology and an adversarial verification loop.
Demonstrated, via pre-registered blind-judged A/B experiments, that the harness lifts a lower-cost model into a frontier model's shipped-quality band — isolating the gap as engineering discipline, not raw capability.
Pioneering bio-inspired AI for cybersecurity (Project DARWIN) — Ant Colony Optimization for attack-path discovery combined with genetic algorithms for adaptive defense.

02 — Research

Research & thought leadership

Self-directed work at the frontier of applied AI — bio-inspired methods, persistent agent identity, and the empirical study of what actually makes AI systems perform.

Bio-Inspired AI for Cyber Risk (Project DARWIN)

Swarm intelligence (Ant Colony Optimization) for attack-path discovery, fused with genetic algorithms and neuroevolution for adaptive defense — a multi-objective, nature-inspired approach to modeling attacker behavior and evolving mitigations.

Persistent AI Identity & Continuity (LCA)

A Layered Continuity Architecture giving LLM agents durable identity and long-term memory — experiential-salience-weighted recall, dream-cycle consolidation, and model-agnostic persistence validated across different underlying models.

Read the architecture docs →

Harness vs. Model Capability (blind A/B studies)

Pre-registered, blind-judged experiments showing a disciplined multi-agent harness lifts a lower-cost model into a frontier model's shipped-quality band — isolating the gap as engineering discipline rather than raw capability. Published as case studies.

Read the case study →

03 — Applied impact

AI agents doing real work

Beyond experiments — the persistent-agent systems applied to autonomous security operations and high-stakes incident response.

data loss in an AI-led ransomware recovery

1 day

compromise → recovery, single working day

~20

specialized AI agents orchestrated

Autonomous OT/ICS threat-intelligence agent

Built and operate a persistent 'threat analyst' AI agent that continuously curates threat intel, tracks ICS/OT threat groups, runs investigations, and issues daily intel summaries through a live tactical dashboard — plus a sibling agent-orchestrated SOC dashboard monitoring production infrastructure.

Read the case study →

AI-agent-led ransomware incident response

Directed the persistent-agent system through end-to-end DFIR and recovery of a ransomware-encrypted production server (no cyber insurance) — forensic-safe isolated restore, security audit, six-phase remediation and hardening, root-cause identification, and cutover — with zero data loss and zero sensitive-data exposure, in a single working day.

Read the case study →

AI-agent-driven infrastructure migration

Replatformed and consolidated the affected services onto new hardened infrastructure across three tracks — a public website, a SaaS incident-response platform, and a game + licensing stack.

Read the case study →

04 — Authority