Selected work

My career has followed a single thread from studying how human memory works at a computational level, through building AI that measurably improved military training outcomes, to delivering enterprise AI projects and creating original intellectual property.

2023 — present

AI Leadership

I currently serve as Chief AI Officer, delivering $3M+ enterprise AI projects while creating privately-owned intellectual property from scratch.

DEER

Deeply Embedded Emotional Resonance — Core IP, 2025

Standard embedding models encode topic, not emotional state. Two people expressing the same frustration about entirely different subjects will land far apart in embedding space, and running an LLM on every utterance for real-time classification is too slow and too expensive.

  • Achieved a 3x improvement in embedding cluster quality over the baseline model.
  • Improved correlation between utterance similarity and emotional state by 73%.
  • Runs at sub-10ms inference with no LLM at runtime and a ~160MB memory footprint.
  • Built ATLAS, a 109,594-point reference mesh using a novel graph-derived interpolation approach to map embedding space in human-readable language.
  • Designed a novel training methodology using natural-language state descriptions as contrastive anchors.
  • A provisional patent is in preparation covering the ATLAS architecture.

I am the sole architect of this system. I designed the training objective, built the canonical corpus and evaluation protocol, developed the ATLAS mesh, and deployed the inference service. Every component was built from scratch.

AI-Native Education Platform

Five IP Systems for At-Risk Learners, 2024

At-risk and non-traditional learners need AI-native education tools that are designed around their needs from the ground up, not traditional systems with a chatbot added on afterwards.

  • I conceived and delivered five distinct platforms: a curriculum authoring system with built-in pedagogical governance, a semantic asset discovery engine, a transcript intelligence system, an enrolment platform with a conversational AI agent, and a learning plan optimiser.
  • The curriculum platform is deployed and operational, with a K-5 launch targeting the 2026-2027 school year.

I owned all five systems from conception through delivery, including architecture, governance model design, AI integration approach, and quality enforcement.

AI Writing Coach

Enterprise AI for a Major Educational Publisher, 2023-2024

One of the world's largest educational publishers needed AI writing coaching integrated into two of their supplemental products. This was their first significant AI project, with serious requirements around security, guardrails, scalability, and cost control.

  • Delivered separate student-facing and teacher-facing AI agents across two product lines.
  • This was the client's first major AI initiative.
  • I personally built cost-modelling tools for the client to help them manage ongoing AI operational costs.

I owned the full engagement from proposal through delivery, covering architecture, agent design, model selection and testing, guardrail implementation, and cost modelling.

Real-Time Legal AI

Prototype, 2026

Real-time analysis of live proceedings against a complete documentary record, a task currently performed manually under extreme time pressure.

  • A provisional patent has been filed covering the core methodology.
  • A multi-million dollar engagement is in negotiation.

I designed the analysis methodology and prototype architecture.

2015 — 2023

Cerego / Memre

I spent eight years as VP Science, where I designed the adaptive learning engine, AI content creation tools, and predictive analytics used by the US military and millions of learners worldwide.

Adaptive Learning Engine

VP Science — 2015-2023

Most training is forgotten within weeks. Traditional approaches cannot personalise to individual memory decay patterns or predict how well someone will retain information in the future.

  • Controlled US military studies showed 27-50% better training retention.
  • Training time was reduced by 33-56% compared to traditional methods.
  • The platform was adopted as part of Booz Allen Hamilton's $937M ET2RC contract with Army Forces Command.
  • It was deployed at F-15 Pilot Training, the Army War College, and Air Force Basic Military Training.
  • In the Air Force assessment, 99% of users passed. My learning analytics predicted retention at R=0.88, significantly outperforming traditional post-tests.
  • I am named inventor on seven granted US patents with additional filings pending, and was first-named inventor on three of those.
  • I built the Smart Create content tools using BERT and early GPT models for automated content generation, well before the current LLM era.

I designed the core learning engine, the predictive analytics system, and the AI content creation tools. I directed all efficacy research with military and university partners and led a research collaboration with the Toronto Lab for Social Neuroscience, which was published in PLoS ONE.

Gates Foundation NGCC

$4M Research Grant — 2016-2019

The Bill and Melinda Gates Foundation funded a multi-year programme to develop personalised courseware for adult learners, with a particular focus on low-income students and those eligible for Pell Grants.

  • I secured and delivered the full $4M research and development project.
  • The work produced a peer-reviewed finding that building foundational knowledge retention improved higher-order understanding.
  • I coordinated with SRI, the Open Education Group, and multiple institutional review boards to design and execute the research.

I directed research efforts for the full grant, designed the methodology, managed institutional partnerships, and delivered the engagement.

2007 — 2014

The Foundation

I completed a PhD and postdoctoral research on the computational modelling of human episodic memory, which laid the scientific foundation for everything that followed.

Episodic Memory Research

PhD Edinburgh, Postdoc UC Davis

Measuring and separating the distinct processes underlying human episodic memory, particularly the success and precision of recollection, required new experimental paradigms and statistical methods.

  • My work has been cited over 500 times across eight publications, and citations continue to grow every year.
  • I published in the Journal of Experimental Psychology, Psychonomic Bulletin and Review, Memory, PLoS ONE, and Behavior Research Methods.
  • The experimental paradigm and statistical framework I developed have been adopted by at least four research groups at other institutions.
  • I co-developed the open-source ROC Toolbox for memory data analysis.

I developed the experimental paradigm, the statistical framework, and the analytical tools. I guided graduate students and collaborated with research groups at Edinburgh, UC Davis, UT Dallas, and the University of Stirling.