What we build
Production AI systems for enterprise infrastructure: live SaaS products, autonomous agents with safety rails, and the network-automation lineage behind them. Everything below is live or public — click through.
Products
Live products
SaaS products NeuralConfig ships and operates in production.
AI-native idea management: semantic search over vector embeddings, capture via MCP, CLI, Siri, and email; Stripe billing on a multi-tenant Cloudflare D1 backend.
Proves: full-stack AI SaaS shipped and operated solo — auth, billing, multi-tenancy, support.
AI voice receptionist for home-services contractors: realtime voice over WebSocket via Durable Objects, Twilio telephony, calendar booking, Stripe subscriptions. $149/month flat.
Proves: realtime voice AI in production with telephony, billing, and compliance surfaces.
Case Studies
AI agent infrastructure
Production systems that put Claude to work against real enterprise APIs and real sales motions.
A reusable code-mode MCP pattern that exposes 1,500+ enterprise-network API operations to Claude through a single tool call — OAuth 2.1 with dynamic client registration and PKCE, running on Cloudflare Workers. Applied across three separate vendor platforms from one architecture.
Proves: sophisticated MCP + agentic tooling design that lets an LLM operate large, real-world API surfaces safely.
A production Claude agent that runs itself: Claude via Cloudflare AI Gateway, Vectorize RAG over a domain corpus, Durable Object workflows, and cron-driven monitoring of dozens of external sources across many upstream platforms.
Proves: production Claude agents with retrieval, scheduled autonomy, and multi-tenant isolation.
An interactive product-demo platform for sales engineers: click-through flows with live-synced speaker notes — flow- and screen-level — driven to a separate presenter window, backed by Cloudflare D1.
Proves: pre-sales enablement built as real software — the SE workflow understood from the inside.
Open Source
Open-source engineering
Public code that shows how the production systems are built.
Autonomous Claude agent that resolves network-support tickets end to end — interprets the ticket, SSHes into the switch, makes the scoped change, verifies, closes. Scope-verification gate and full audit trail.
AI pipeline rating 1,200+ feature requests across 9 market verticals — local LLMs via Ollama with a tiered model cascade, three-stage duplicate detection, Salesforce revenue correlation.
IoT fingerprinting fusing 7 evidence sources with weighted confidence scoring; RADIUS integration moves devices quarantine → classify → production VLAN.
Lineage
Why it holds up in production
Twenty-plus years of enterprise network operations and automation is why the AI systems above hold up in production — the constraints (auth, audit, failure modes, compliance) were learned the hard way first.
Programs