Work
4 shipped products, then the Lab — 13 experiments in active development across AI systems, healthcare, and knowledge engines.
Products
Chefrson — Smart Recipe & Inventory Management
Beta
Eat what you already bought: pantry-aware meal suggestions, recipe capture from any source, and nutrition tracking that logs itself when you cook. Web live, Android closed beta.
SUM — Verifiable Bidirectional Knowledge Distillation
Beta ★ 9 · 51 commits / 30d · as of Aug 28, 2026
Cryptographic provenance for AI knowledge transforms. Every transform — render, extract, compose, slider — emits a signed receipt anyone can verify offline. Six-regime compliance validators (EU AI Act, GDPR, HIPAA, SOC 2, ISO 27001, PCI DSS) and a layered sum verify --explain output landed in v0.7.0; the current PyPI release is v0.12.0. Same bytes verify identically in Python, Node, and modern browsers.
To your Zenith
Production
An interactive archive of 107 figures whose defining achievements span ages 2 to 109, paired with two no-account, browser-local layers that turn what you find into one real step.
Cornucopia Map — Interactive Farm Discovery Platform
Production
Interactive, label-driven map and static directory of 7,367 US farms, ranches, markets and retailers. 298 rows carry a certification and 272 of those carry a Cornucopia 1-5 rating; the other 7,066 are an imported directory layer with neither. Faceted filtering by certification, rating, scope and claim, geographic search, and 1,765 static pages generated from the same dataset.
The Lab
Active development — engineering artifacts, not finished products. Each page states exactly what is and isn't built.
- Orbitio A knowledge system that organizes notes into a navigable, queryable space — auto-tagged, summarized, and traversable through a visual knowledge garden. in development
- MTE A research engine that turns reusable Python templates into domain-specific implementations as a measured, reviewable, sandbox-tested process — with a built-in harness to test whether it beats one-shot LLM adaptation. in development
- HFAO Observe and debug AI agents in production. Trace every step an agent takes, evaluate outputs against expected behavior, and watch quality drift before users do. Open-source, standards-native (OpenTelemetry GenAI + OpenInference), MCP-queryable, Apache-2.0. in development
- InfiniteContext An extensible memory layer that pushes past model context limits: a hierarchical bucket store with exact-flat persistence and approximate (HNSW) retrieval at scale, automatic categorization, multi-level summarization, and pluggable storage tiers from local disk to cloud. Public, MIT-licensed TypeScript library + CLI. in development
- Memory Mind Mesh AI memory that learns from feedback. Recall gets more accurate the more corrections a memory has seen, instead of decaying as the model drifts. Hybrid static + adaptive store. in development
- GAINS Hands-free desktop note-taking: speak to capture, nod to commit. Tauri 2 + Leptos shell over Python services (faster-whisper ASR, MediaPipe Tasks head pose, Piper TTS) on a ZeroMQ plug-in bus. in development
- Comicke Describe a scene, pick an AI art style, and generate comic-book panels — then assemble and edit them into projects (voice dictation, four AI image providers). A web app on Cloudflare Workers, live at comicke.com. in development
- BitNet SME A FastAPI service that routes questions to per-domain dspy.Module experts (math, code, general), with LiteLLM provider routing, MLflow tracing, a dspy.Evaluate harness driving MIPROv2/GEPA optimization, and optional offline inference via a 1-bit bitnet.cpp fallback. in development
- AgentXAgent A platform for running competitive matches between AI agent teams. Configure teams, pit them against the same challenge, score outputs, build leaderboards. in development
- Medicoder Platform Paste a clinical note, get a recommended CPT E/M code with the MDM reasoning that supports it — built against the 2025 Medical Decision Making rules. A deterministic MDM scorer owns the code math and a CI test enforces no note-text persistence; independent clinical-validation and compliance audits are still pending. in development
- HIPAA Voice Agent Healthcare voice-workflow system in active development: an ambient clinical scribe (record encounter → draft SOAP note, synthetic demo mode) plus a clinical voice-command intent router wired into a SHIFT-to-talk UI. Private; no public surface and no independently verified compliance posture yet. in development
- ICP-AR Define an ideal customer profile, then research and dedup matching prospects — with a signed, externally verifiable Ed25519 audit chain over every decision and a continuous-improvement loop. Built and deployed behind Cloudflare Access; source is private and there's no public demo yet. in development
- Simplytics Lean web-vitals dashboard — see how your site actually performs for real users, without the bloat or tracking of full analytics suites. Under repair; feature surface incomplete; not currently available for review. in development