Addison Smith
Controls Application Engineer at Mantis Innovation, bridging mechatronics and intelligent software. Building at the intersection of automation and AI.
Engineer. Developer. Builder.
I integrate BACnet networks at 8am and train neural networks at 8pm. I bridge the gap between industrial hardware and intelligent software.
Controls Engineering
PLC troubleshooting, system integration, industrial wiring, and automation R&D.
AI & Machine Learning
Computer vision, model training, data pipelines, and agentic systems.
Software Development
Python, JavaScript, SQL, web apps, and database design.
My goal is to design the next generation of industrial systems where automation, AI, and real-world engineering come together.
Tech Stack
Experience & Education
Work Experience
Controls Application Engineer
Mantis Innovation
Remote
Building Automation Controls
- Design and application engineering for Mantis BMS projects
- Niagara framework work across hyperscale data center building systems
- Ignition for supervisory control and visualization
- BACnet and Modbus integration across building equipment and controllers
Controls Engineering Technician
Daedalus Industrial
Controls Engineering
- Programmed and commissioned PLC systems in Studio 5000 Logix Designer (CompactLogix/ControlLogix), including sequencer logic, PIDE loops, valve control, and pump control routines
- Designed and deployed HMI interfaces in FactoryTalk View for operator monitoring and process control in regulated manufacturing environments
- Executed wiring, setup, and testing of control panels for automation and R&D projects, following UL508A and NEC safety standards
- Commissioned systems on site, including tank farm automation using Ethernet/IP and Modbus TCP protocols
Machine Learning
- Built and deployed computer vision inspection systems for real-time defect detection in manufacturing
- Developed an L5X generator that programmatically creates Rockwell ladder logic by ingesting existing L5X project files and building new programs from templates for direct import into Studio 5000
- Designed a multi-agent AI system (CrewAI/Ollama) for engineering automation, including vision analysis, code generation, and task orchestration across specialized teams
- Supported data collection, labeling, and model training to improve detection accuracy across production lines
Information Technology
- Provided general IT support for control systems and connected devices, ensuring smooth integration and network functionality
Electrical Technician
AIRSYS USA
- R&D support for cooling systems used in chip and data center applications
- Build and wire 110V–480V control panels
- Read schematics, troubleshoot systems, and configure sensors
- Collaborate with engineers on prototype testing and updates
Wiring Technician
Vertiv
- Wired industrial switchgear systems following detailed schematics and NEC safety codes
- Conducted quality checks to ensure accuracy and reliability of assemblies
Education
BS Information Technology
Software Development & Coding, AI/ML and Data Analytics track
ECPI University
Bachelor of Science in Information Technology with a concentration in AI/ML and Data Analytics. Eighty-five of one hundred twenty credits earned at a 4.0, on track for February 2027 completion.
AS Mechatronics, Robotics & Automation Engineering
Piedmont Technical College
Graduated with 3.8 GPA. Hands-on training in electrical, mechanical, and automation systems.
Things I've Built
BMS Alarm Triage Agent
A building automation alarm queue goes in. A short ranked list of what an engineer should look at comes out, each item backed by trend evidence and the named rule behind its rank. The pipeline is mostly deterministic, so every result is traceable. Claude Haiku writes the two-sentence explanations at build time, and they pass the same guardrails as everything else. Read-only, with no write path to any building system.
548 alarm events down to 18 items to review on the holdout set, with all 15 target escalations captured. The evaluation data is synthetic, so this shows the pipeline and test harness working end to end, not proven performance on a real building.
Distech GFX Tooling
Tooling for authoring Distech GFX programs as text instead of function blocks by hand.
Niagara Station Tooling
Tooling for working with Niagara station configurations offline instead of station by station in the engineering tool.
Ignition Site Pack Generator
Generates an Ignition Perspective site from spreadsheet equipment lists. Produces the UDT library, tag instances, parameterized detail views, navigation, scoped alarms, and a tenant billing chain.
Smith_Agentic
A modular multi-agent framework for engineering workflows, built as four components: a Python/CrewAI backend with twenty specialized agents across four crews; a React UI with live WebSocket streaming, 18 pre-built task templates across Code, PLC, HMI, Vision, Safety, and Ops, and an offline demo mode; an MCP server that exposes crew runs as tools for Claude Code and Claude Desktop; and a Rust CLI for driving the backend from the terminal.
Twenty specialized agents. Four crews. One framework that turns engineering goals into finished deliverables without touching the cloud.
Argus
Industrial anomaly detection platform. Monitors live Modbus register data using statistical analysis, generates plain-English diagnoses via local LLM, and fires alerts before problems become failures.
Watches your Modbus registers around the clock, catches drift before it becomes failure, and tells you exactly what's wrong in plain English.
Vision_Inspect
A local shop-floor visual process monitoring system using Visual Language Models via Ollama for real-time inspection of manufacturing processes. Sees everything, touches nothing, reports everything.
Local VLM-powered visual inspection for manufacturing.
ModBridge
Modbus TCP utility that polls live register values, serves them as a REST API, logs to CSV, and runs a built-in fake device for testing. Ships in both Rust and Python with identical feature sets.
Industrial field data, turned into a clean REST API, in one binary with no runtime required.
Power Dashboard
Real-time Electrical Power Monitoring System dashboard with simulated Modbus data collection, ISA 18.2 alarm management, interactive one-line diagrams, and historical trending. Built for data center and industrial facility monitoring.
Full-stack EPMS with real-time WebSocket data, 8 simulated meters, and ISA 18.2 alarm engine.
Other Projects (2)
Chalk
An AI football intelligence platform built for coaches. Three modules cover film analysis, recruiting evaluation, and game plan recommendations, powered by recognition, analytics, and recommendation intelligence layers.
Three modules. One platform. Film, recruiting, and game planning intelligence built for coaches who think in systems.
Sift
A camera-first personal decision engine. Point your phone at any product, restaurant, or item and get an instant AI-powered score, verdict, and ranked alternatives pulled from real reviews and local options. Built with preference learning so results get smarter over time.
Point. Scan. Decide. AI-powered verdicts on any product, restaurant, or item in seconds.
Certifications & Achievements
Building Agentic AI Applications with Large Language Models
NVIDIA Deep Learning Institute
VerifyExpert Panel Contributor, Data Center Cooling
Mordor Intelligence
2025Recognized as a member of their Expert Panel Group for contributing subject matter expertise and strategic insights to the Data Center Cooling market report.
VerifyFoundational Certifications (6)
From the Blog
From Tech to Engineer
Five roles and no plan. On stepping into whatever gap was open, and finding out later that it was a career.
Because I Saw It
On recognizing the right opportunity when it appears -- and what it takes to actually act on it.
From Story to Systems: Breaking Down My Portfolio and Power Dashboard
A technical breakdown of two projects. No fluff. Just how they work.
Let's Connect
I'm always interested in connecting with people working at the intersection of industrial automation and intelligent software. Whether you want to talk about controls engineering, AI/ML, or potential opportunities, feel free to reach out. I typically respond within 24 hours.