Building end-to-end — from self-hosted AI knowledge bases to industrial predictive maintenance systems.
A live-deployed, browser-based dashboard that monitors up to four industrial weight indicators simultaneously — rail scales and truck scales — over TCP, presenting real-time readings on a facility display. Built for an industrial client and running in production.
Production RAG chatbot for industrial scale and controls documentation. Ollama + Chroma vector DB + Open WebUI. Answers equipment-specific questions by retrieving from actual manuals.
Multi-channel AI triage for support tickets — SMS, email, and web. LLM classifies, prioritizes, and routes issues. Integrates with Twilio and the knowledge base.
Customer-facing portal with AI dispatch logic — auto-assigns technicians based on issue type, location, and expertise. Customer NLP → scheduling engine → tech notification.
Computer vision pipeline that reads paper forms, extracts fields via OCR + LLM parsing, and auto-fills digital systems. Targets industrial inspection and compliance paperwork.
Hardware-in-the-loop bridge connecting real PLC/sensor hardware to AI pipelines via MQTT and HTTP. Raspberry Pi + Arduino sensor simulator feeding live data to ML models.
ML pipeline ingesting load cell readings, temperature, drift, and service history to predict equipment failure 7–30 days out. InfluxDB time-series + Isolation Forest + failure regression. Auto-triggers service dispatch.
Real-time vision inspection for manufacturing lines. Camera feed → CV model → pass/fail decision → PLC output signal to reject mechanism. Bridges AI inference and hard real-time control.
Three dedicated controls projects demonstrating PLC programming, HMI design, and industrial commissioning — the domain expertise beneath the AI work.