Lab
Applied experiments and deployed systems.
Beyond the flagship products, I build and deploy smaller systems that explore specific questions about AI automation, data integration, and real-world application. Each project teaches something that informs the larger work.
Article Intelligence →
What It Does
A contextual AI system embedded across this website that answers questions about article content in real-time. The system logs interactions, categorizes questions using LLMs, identifies content gaps, and feeds this intelligence back into content generation. Users can rate responses, creating a continuous feedback loop for quality improvement.
What It Taught
Demonstrated how AI systems can learn from user behavior at the edge. The real value is not in answering questions but in understanding what questions people ask, which reveals gaps in existing content and guides future creation.
Curriculum-Aware Learning Agent (Samson)
What It Does
A persistent AI learning agent designed to support continuous, goal-driven education through conversation. The system aligns freeform dialogue to a structured curriculum, validates prerequisites before advancing, tracks mastery with temporal decay, and verifies understanding through critical reasoning rather than recall-based testing. Learning state, progress, and memory persist across sessions, enabling the agent to remember how understanding developed over time.
What It Taught
Demonstrated that learning systems become significantly more effective when memory, progress, and verification are treated as first-class architectural concerns rather than session artifacts. Genuine learning emerges from experience-grounded interaction: remembering how understanding developed, not just what content was delivered. This project highlights the importance of episodic traces, semantic consolidation, and curriculum-aware adaptation for AI agents intended to support long-term learning and identity persistence.
Adaptive Learning Platform
What It Does
An interactive video-based education system designed to reshape how students learn. The platform dynamically adjusts content delivery based on individual learning patterns, moving away from traditional multiple-choice assessments toward more meaningful comprehension measures. Currently in beta with a major university partner.
What It Taught
Education technology has focused on content delivery, not learning adaptation. The insight is that engagement patterns, pause points, and replay behavior reveal more about comprehension than test scores. Similar to how Article Intelligence adapts to user questions, this system adapts to how students actually learn.
AI Backyard Redesign Tool ↗
What It Does
A computer vision tool that transforms backyard photos into professional landscape design concepts. Homeowners upload a photo of their current yard, select a design style (Modern, Japanese, Mediterranean, etc.), and receive an AI-generated visualization of the potential transformation. The tool serves as a zero-friction lead generation system for a landscaping business.
What It Taught
Demonstrated how generative AI can reduce friction in service businesses. Traditional landscaping consultations require scheduling, site visits, and manual mockups. This tool lets potential customers explore possibilities before any commitment, qualifying leads while providing immediate value.
Mentor Intelligence Program
What It Does
An AI system designed to support mentorship programs and student engagement. The system tracks interactions, suggests discussion topics, identifies students who may need additional support, and helps mentors maintain meaningful connections at scale.
What It Taught
Explored the challenges of building AI that supports human relationships without replacing them. The goal is augmentation: helping mentors be more effective, not automating mentorship.
IT Sustainability Finder
What It Does
A tool that maps technology assets to sustainability metrics. Organizations can input their IT infrastructure and receive analysis of energy consumption patterns, carbon footprint estimates, and recommendations for more sustainable alternatives.
What It Taught
Highlighted the value of connecting disparate data sources. Sustainability data exists but is fragmented across vendors, certifications, and research. Aggregation and standardization create new insights.
Live Trading Models
What It Does
Machine learning models that analyze market data and generate probabilistic signals for trading decisions. The system processes real-time data streams, identifies patterns, and outputs risk-adjusted recommendations.
What It Taught
Reinforced the importance of uncertainty quantification. Markets are noisy; the value is not in predicting outcomes but in accurately estimating probabilities and managing risk accordingly.
The Lab Philosophy
The lab exists because ideas need to be tested. Reading papers and thinking about systems is valuable, but building and deploying reveals constraints that theory misses.
Each project here started with a specific question or opportunity. Some became ongoing systems that serve real users. Others were experiments that answered their question and concluded. Both types are valuable.
The projects tend to share common threads: integrating multiple AI capabilities into coherent workflows, connecting disparate data sources, and finding ways to make AI systems genuinely useful rather than impressive demos.
Have a project in mind?
If you have an applied AI challenge that might benefit from this kind of experimental approach, I am interested in conversations about potential collaboration.