◇
Query-Driven Memory
Persistent AI memory with vector similarity retrieval. Memories reinforce with use, decay over time, and resurface automatically when relevant.
React · FastAPI · Qdrant · Ollama
⌫
Model Security Scanner
Automated LLM vulnerability testing across 10 attack categories — prompt injection, jailbreaking, data extraction. Scans local and cloud models.
React · FastAPI · Ollama · Claude API
△
Local RAG System
Private document-grounded chat. Upload documents, ask questions, get sourced answers. Fully local — no API keys, no cloud dependencies.
Docker · Qdrant · Ollama · Open WebUI
◈
AI Inference Cost Calculator
Total cost of ownership for GPU builds vs cloud inference. Break-even analysis across 6 hardware tiers over 36 months.
React · FastAPI · Recharts
⬡
Datacenter Demand Simulator
Monte Carlo capacity forecasting across compute, storage, power, and cooling. AI-powered scenario analysis with decision timelines.
React · FastAPI · Ollama · Recharts
▣
DC Optimization & Valuation
Datacenter investment analysis for PE/VC. Financial modeling, M&A valuation, and deal memos across 14 global markets.
React · FastAPI · Recharts
◖
AssetPulse
Transit asset management platform built on public MBTA data. FTA condition scoring, capital planning, and state-of-good-repair analysis across 200+ assets.
React · FastAPI · SQLite · Recharts
▰
CIO in the AI World
12-lecture executive education course. Strategy frameworks, live demos, and implementation playbooks for enterprise AI adoption.
React · FastAPI · HTML5 Slides
○
Enterprise Production Platform
Full-stack model health scoring with 31 automated metrics, 209 passing tests, C# cloud extraction pipeline, and 15 dashboard pages.
React · FastAPI · PostgreSQL · C# · Autodesk APS
How I Build
In December 2025, I started an experiment. I'd spent 40 years leading technology organizations — managing teams, budgets, and enterprise infrastructure. I understood deeply what needed to be built. But I'd never been the one writing the code.
Claude Code changed that equation. In 10 weeks, I built everything on this page. Not by writing code full-time — by directing Claude in focused sessions between meetings, conference calls, and life. I'd describe what I needed, kick off a build, and come back to working software. The traditional estimate for this scope: 2–3 specialists, 17 months, $250,000+.
10 wks
Calendar time
vs 17 months traditional
~$500
Tool cost
vs $250K+ traditional
1
Person
vs 2–3 specialists
43,640
Lines of code
7 languages · 12 apps
Claude Code doesn't replace developers. It eliminates the translation layer between "the person who knows what needs to be built" and "the person who knows how to build it." When those are the same person — augmented by AI — the result is transformative.
Every project on this page was built with
Claude Code (Anthropic).