Factory AI Platform ● Live Build
End-to-end industrial intelligence — from raw PLC signals to AI-driven root cause analysis.
What it is
A unified, real-time system that brings together disconnected factory data and enables AI-driven intelligence across every layer of the production floor.
The Problem — Three Isolated Silos
The Solution — 5 Layers
Tech Stack
| Layer | Tech | Role |
|---|---|---|
| Edge | Rust | Mock PLC (20 samples/sec) |
| Streaming | NATS JetStream | Low-latency event routing |
| Enrichment | Go | Join PLC data + MES state |
| Analytics DB | ClickHouse | Time-series storage |
| Anomaly Detection | Python / FastAPI | Statistical + ML signals |
| AI Reasoning | LLM + Extracta AI | Context-aware root cause analysis |
| Dashboard | React / TypeScript | Real-time visualization |
Business Value
| Problem | Outcome |
|---|---|
| Quality escapes | Catch anomalies before parts ship — 5–10% fewer defects |
| Downtime | Predict failures before they happen — 20–30% less downtime |
| Root cause mystery | AI explains what went wrong in seconds — 50% faster resolution |
| Bottlenecks unknown | OEE trends show where to optimize — 10–15% throughput gain |
What This Demonstrates
PLCs, MES, OEE, real production constraints
Event streaming, NATS, concurrent processing
Observability, graceful degradation, backpressure
Rust + Go + Python + React + databases
ML inference in a real systems context, not notebooks
Current Status
-
✅ Complete — Streaming, enrichment, storage, dashboard
NATS JetStream event routing · Go enrichment service · ClickHouse time-series · React real-time dashboard
-
⏳ In Progress — Observability & anomaly detection
Prometheus / Grafana metrics · Statistical anomaly signals · ML-based detection layer
-
🔜 Planned — GraphQL, Kubernetes, Extracta AI
GraphQL query layer · Container orchestration · LLM-powered root cause analysis via Extracta AI