Systems I've built, broken down.
Each one links to a deeper write-up — the architecture, how the data's captured, and the part that was actually interesting to solve. The two active R&D efforts share the architecture that is ready to discuss today.
GCPS Azure Local Operations Platform
Not a normal monitoring dashboard. A full-stack platform with a built-in AI assistant (TanStack AI), the Forge automation runner, and an evidence-first model over Azure Local (formerly Azure Stack HCI) — PowerShell collectors, SQL Server as a processing engine, typed APIs, and a desktop-grade React dashboard.
The Forge
Run real PowerShell from a browser — safely. SQL-backed job queues, command manifests, runner heartbeats, artifacts, cancellation, and a full audit trail. Built into GCPS, important enough to stand on its own.
Ask ALDHI
An AI assistant with its hands on the evidence — 39 typed tools over every data source the platform collects, plus a sandboxed code mode, with hard guardrails: no raw SQL from the model, no secrets, row caps, approval-gated writes. Root cause in minutes, not hours.
AlphaForge — stock automation & AI Lab
An evidence-capturing market-automation platform with an AI Lab and a risk engine. The current write-up shares the evidence-first architecture while the code remains closed.
design-doctor
A configurable design-quality compiler for React, shadcn/ui, Tailwind v4, motion, and accessibility — scores UI, governs motion drift, ships as an installable agent skill.
Azure Local Redeployment Console
A scrappy console tool I built in about a day: menus, per-site progress, and an audit log of every command run — making sure every fiddly pre- and post-install step happened in order. That's how one person knocked out 96 migrations in 10 days.
Fintech Analytics & Automation
Next.js analytics plus a Python Azure Functions document pipeline — OCR, invoice linking, ACH/bank reconciliation, SQL metrics, and executive risk views.
Fiber Optic Project Management
Next.js 16 telecom ops platform — KMZ/KML route maps, Azure Blob documents, RBAC, PO/invoice/payment workflows, and payroll reserve.
Video Frame Analysis Pipeline
Python CLI that scores every frame with pyiqa/PyTorch, stores rankings in SQLite, and exports traceable stills through hardened FFmpeg paths.
Happy to walk any of these end-to-end.
The case studies cover the architecture and engineering decisions; a live tour adds the operational context that cannot be published.