forward deployed engineer · full-stack builder

Aaron Simo.
I go where the work happens and build what makes it run.

I'm the on-site technical lead for the US operation of a German photobooth manufacturer. A few thousand devices a month pass through my team, and I build the tools, test rigs, networks, and dashboards they run on, right now including the IT buildout of a new building. I also build and ship full-stack products on my own, and I use AI agents every day to do it. Before all that I spent seven years fixing avionics in the Navy. The thread is simple: I like making things work reliably, and building the thing that makes the next person's job easier.

See the KRUU work View résumé ↗
aasimo13@gmail.com · 586-344-1284 · LinkedIn ↗ · Clinton Township, MI
01 In the field · KRUU US

My day job is the closest thing to forward deployed engineering I know of. I'm the person on site who makes the technology work for a real operation, working with engineering and IT at headquarters in Germany. Here's a slice of what I've built there.

3D model of the new KRUU US warehouse showing shipping shelves, check and clean rows, conveyors, racks, and building steel, merged with phone scans of the real floor

The warehouse planner's 3D view: CAD layout merged with phone scans of the empty building.

Warehouse planner, 2D + 3D

new facility buildout · used by the KRUU US team
in use

When we took a new building, I built the tool we planned it in. It's one page with a 2D floor plan and a 3D view, built from the architect's PDF and LiDAR scans of the empty building. It lays out shipping shelves, check and clean rows, conveyors, racks, and building steel, and it runs the power plan with panel schedules and load checks plus a data port plan, with PDF, PNG, and Excel exports. It also loads the conveyor vendor's own CAD model (8,990 parts from their STEP file) so we could check their layout against ours before anything was installed. I packaged it so our German site can rebuild it for their warehouse.

three.js LiDAR scans · glTF STEP to mesh PDF floor plan parsing Power & load calcs Python tooling

Facility IT buildout

I own IT and workshop for the move: internet cutover, network and switching to every fulfillment shelf, bringing the building's inherited access control and camera systems onto our network, and the shelving layout, power distribution, and cabling.

MiLa test station

Raspberry Pi pass/fail rig for USB microphone units. Plays a frequency sweep, records it back, and validates the capture in about five seconds, with auto plug-in triggering, LED and beep output, and self-calibration.

Printer test rigs code ↗

Plug in a DNP or Canon SELPHY printer and the rig detects it over USB and runs a 10 image QC suite (color, resolution, text, edge bleed) with a live web dashboard. Headless, with a one command reset for the floor.

DNP print server code ↗

Pi based photo print station: drag photos into a web page or a shared folder and they print at 4x6, 4x4, or cut strips. One command installer that detects the printer and user on its own.

Photobooth Test Box

A diagnostic box for the fleet that checks the 24V supply, USB voltages, the pedal switch, and cable continuity, so a tech can find a fault in minutes instead of swapping parts.

Floor dashboards & parts

Metabase dashboards on Pi kiosk screens and a mobile fulfillment widget for live throughput and first pass yield. CAD and 3D printed parts for the fleet, like a retention clip for the DNP printers.

the centerpiece · a living experiment

Operant

A persistent reinforcement-learning organism, shaped forever by everyone who visits.

A small learning mind lives on a server and never stops. It moves, it learns, and every visitor who has ever been here has changed it a little, permanently, through reward and punishment. Nothing is ever undone. It's part art installation and part engineering piece, inspired by Sun Yuan & Peng Yu's Can't Help Myself and by simulation theory. Built as a TypeScript monorepo: a framework-agnostic Q-learning core, an always-on simulation host that owns and persists the single agent, and a react-three-fiber 3D view for observers, with the learning core kept strictly decoupled from rendering.

Enter Operant → View the code ↗
TypeScript Reinforcement learning · Q-learning Node.js React · react-three-fiber pnpm monorepo Railway · GitHub Pages
02 Projects
live · web + iOS

A family-management app with AI-assisted meal planning, smart shopping lists, shared finances, scheduling, and chore gamification. Shipped on the web and the App Store, built and deployed solo: a web frontend, a Supabase Postgres backend with row-level security and hardening migrations, edge functions on a GitHub Actions pipeline, and an iOS build with Apple in-app purchases. It's my biggest project, and I designed, built, deployed, and have maintained all of it.

Supabase / Postgres Row-Level Security Edge Functions iOS · App Store Apple IAP GitHub Actions CI/CD Vercel AI-assisted features

Argus

fully-local AI home monitor · self-hosted
AI / inference

A fully-local, AI-powered home-monitoring and face-recognition system for Home Assistant. No cloud, and everything runs on your own hardware. It's three cooperating services: a Watcher that snapshots the home's state every few minutes and asks an LLM "is anything actually wrong?" (a hardened, high-signal prompt with an urgent-push vs. daily-digest policy); a Recognizer running InsightFace on CPU, exposing a CodeProject.AI-compatible API to detect and identify faces on-device; and a Console that clusters unknown faces by similarity so you name a group once and it learns that person from then on. Pure logic is unit-tested; the container and glue layers are dry-run-verified.

Python LLM reasoning loop InsightFace On-device / CPU inference Face recognition Home Assistant · AppDaemon Docker / Compose Flask API Pytest

gutter.guru

gutter.guru ↗
live business

A website and CRM I built for Guys Gutters LLC to manage customers, jobs, and scheduling. I designed and deployed it for a real, operating business. It's a purpose-built tool that replaces spreadsheets and scattered notes with one system the company actually runs on.

Full-stack web app CRM Scheduling Custom domain
03 Also built
Kova self-hosted agent
A self-hosted AI agent on Open WebUI wired to the Anthropic API, deployed on Render.
AI-assisted research side projects
I use Claude agents to take open questions from idea to checked result: lit review, code, proofs, and stress tests. Write-ups coming here.
Experiments prototypes
A steady stream of prototypes. The way I learn a new tool is by shipping something with it.
04 About
85
technicians led
(Navy avionics)
$79M
cost savings delivered fleet-wide
~2,500/mo
devices fulfilled, peaking at 1,000/wk
$1–5M
device inventory managed

I lead technical teams and build the systems that help them move faster. My background is hands-on electronics, and most of my work has been keeping high-volume repair operations running, coordinating fulfillment, managing a few million dollars of inventory, and writing the internal tools that cut cost and keep things reliable.

Right now I lead workshop operations at KRUU US, a German-owned photobooth manufacturer, and I'm the US point of contact for all IT. I direct a team of up to eight people, we move around 2,500 devices a month, and I've built the repair workflows, test fixtures, and dashboards that raised our first-pass yield and gave everyone a live view of throughput. I'm also running the IT buildout of our new building. Before KRUU I spent seven years in the U.S. Navy as an aviation electronics technician and production supervisor, where I supervised 85 technicians keeping F/A-18E/F avionics running across 15 squadrons. Our maintenance work helped save $79M while holding 100% fleet availability.

It's all the same instinct, really. Whether it's a test rig that checks a device in five seconds, a dashboard that flags a bottleneck, or a SaaS backend that has to stay up for a paying customer, I care about getting things to work and building the thing that makes the next person's job easier. I also build full-stack software on my own, and I use AI tools like Claude, GPT, and Gemini every day in my technical work.

now
Lead, US Workshop Operations @ KRUU US Inc
building
KRUU warehouse planner, HomeForge (web + iOS), Operant, Argus, gutter.guru
background
U.S. Navy avionics (AT2) · Six Sigma · Secret clearance (inactive)
stack
Python · TypeScript / React · three.js · LLMs, agents & MCP · Docker · Supabase / Postgres · Firebase · Vercel · iOS · Raspberry Pi · networking
05 Contact

Open to forward deployed engineering and applied AI roles, and engineering leadership. Based in Clinton Township, MI. The fastest way to reach me:

aasimo13@gmail.com 586-344-1284 LinkedIn ↗ View résumé ↗ Download résumé ↓