Advanced Process Technologies
Industrial AI co-pilot for dairy and food processing.
Agentic AI that understands your plant -- documents, sensors, batch records, and PLC programs -- and answers engineering questions with cited evidence from your own operational data.
Live on a production ACV01 Automated Cheese Vat
What is GUS.ai
GUS.ai is an agentic AI system purpose-built for industrial food and dairy processing plants. It combines three knowledge sources that plant engineers have never been able to query together: your document library, your live sensor data, and your PLC logic.
Ask GUS.ai why a set time exceeded spec. It will retrieve the SOP target from the recipe document, pull the actual step timing from the telemetry data, and tell you exactly how far off you ran -- with citations for both sources.
Faster troubleshooting for plant operations. A complete record for management. Evidence-backed and traceable answers for compliance.
Why did the set time run long on last night's batch?
Set time target per recipe spec: 3:00:00. Actual elapsed from telemetry: 5:10:22 -- +72% over spec. Vat temperature held at 88.4F (within range). pH trending data shows coagulation onset delayed by ~18 min relative to historical median. SOP Section 4.3 recommends verifying rennet dose rate if set time exceeds target by more than 30 min.
Sources: Master Recipe Rev B | ACV01 Telemetry | Tech Services Trial #7
Capabilities
11 AI tools, 40+ API endpoints, 5 communication channels -- built to answer the questions plant engineers actually ask.
Semantic search across SOPs, P&IDs, equipment manuals, recipe specs, parts lists, and commissioning reports. Every answer cites the exact source, page, and section.
Real-time sensor data from your process. Current values, time-series trends, and threshold monitoring across every instrumented point in your plant.
Step timing analysis, batch history comparison, and spec-vs-actual deviation detection. Automatic batch boundary detection without MES -- discovered from raw tag transitions.
Parsed L5X programs -- 88 UDTs, 2,027 controller tags, and 480 alarm definitions indexed as 334 RAG chunks. Ask about any rung, any tag, any alarm in natural language.
Anomaly detection using MAD-based z-scores with coincident spike grouping. Multi-tag Pearson correlation analysis. Impact propagation, root cause analysis, and single-point-of-failure detection across a 30-node dependency graph.
Ask GUS.ai from the web app, by email, by text message, by phone call, or by Telegram. Get answers wherever you are -- on the plant floor, in the office, or on the road. Same AI pipeline on every channel.
Automatic email notifications on batch start/end and CIP transitions. Heartbeat monitor sends Telegram alerts for alarms, step changes, and abnormal conditions. System status monitor with incident/recovery emails.
Append-only audit trail of every query, tool call, and response. Daily digest emails with per-user breakdowns. Golden question management for QC evaluation. Full CSV export. Clerk JWT authentication with invite-only access control.
Pump curves, heat exchanger specs, RTD calibration data, flowmeter configurations, automated valves, and level transmitters -- any component or instrument in your process.
Answers stream in real time via Server-Sent Events -- no waiting for the full response. Extended thinking display shows the AI's reasoning process as it works through your question, tool by tool.
Visualization
Four interactive 3D graphs, a live digital twin dashboard, and a causal reasoning panel -- all built into the same interface.
Interactive 3D Graphs
Four domain-specific 3D graphs rendered with D3-force physics. Navigate 900+ code entities, 679 RAG document chunks, markdown documentation structure, and process dependency relationships. Click any node for AI-generated purpose descriptions. Built-in search, type filtering, tour mode, and ambient audio.
Live Monitoring
Real-time ACV01 vat overlay with positioned sensor callouts, radar chart performance view, gauge cards, health score, active alerts, and mini sparklines. Make Room view shows all vats at once with 10-second refresh.
Decision Support
16 pre-built questions across four analysis categories: impact propagation, root cause analysis, single-point-of-failure detection, and process health. One click from question to AI-powered answer backed by a 30-node, 55-edge dependency graph.
Visual Polish
WebGL ferrofluid orb reacts to AI state (idle, listening, thinking, speaking). 10,000-particle intro animation. Neural network canvas background. Streaming markdown with progressive rendering. Evidence panel with Grafana deep-links for answer verification. 4K kiosk mode for trade show displays.
AI Architecture
GUS uses five specialized AI models from two providers -- each chosen for a specific task, balancing reasoning depth, speed, and cost.
Deep Reasoning & Graph Intelligence
Powers interactive code graph, knowledge graph, and docs graph conversations -- deep structural analysis and architectural reasoning across the entire codebase.
Highest reasoningCore Agent Brain
The primary orchestration model. Manages an 11-tool agentic loop with extended thinking -- deciding which documents to search, which tags to query, and how to synthesize evidence into cited answers.
11-tool orchestrationReal-Time Descriptions
Generates instant purpose descriptions for graph nodes on click. Optimized for sub-second latency with server-side response caching.
Low latencyAutomatic GPT-5.2 failover for 99.9% uptime
Voice-to-Text
Transcribes spoken questions into text for the agent pipeline. Operators can ask GUS questions hands-free from the plant floor via voice, phone, or kiosk microphone.
Speech recognitionText-to-Speech
Reads answers aloud with natural voice synthesis. Multi-provider support (OpenAI, ElevenLabs, Mistral). Enables hands-free operation during production runs and equipment inspections.
Voice outputRoadmap
GUS.ai is actively evolving. Here's what we're building toward -- each milestone informed by real plant floor feedback.
Know what's coming before it happens. Statistical extrapolation, learned batch models, and physics surrogates for proactive process advisories.
In researchDedicated AI agents for cheese yield optimization, CIP validation, and preventive maintenance -- each with deep domain knowledge and scoped tool access.
PlannedConnects to any PLC through ADX -- no historian licensing required. Open-source stack, vendor-agnostic. Multi-historian adapters for PI, InfluxDB, Ignition, FactoryTalk.
PlannedAsk GUS from the plant floor. PWA-first with native iOS and Android wrappers for push notifications and offline caching.
PlannedRun GUS entirely on-premises with local LLMs -- no data leaves the plant network. Docker Compose, edge inference, zero cloud dependency.
PlannedAuto-built knowledge graph from 679 chunks via LLM extraction. Cross-modal linking between tags, PLC definitions, alarms, and documents. Upgrades causal reasoning from 30 to 500+ nodes.
DesignedBuilt by APT
Advanced Process Technologies (APT) is a 100% employee-owned process engineering and equipment manufacturing company headquartered in Cokato, Minnesota. Since 2000, APT has designed, fabricated, programmed, installed, and commissioned processing systems for dairy and food plants across North America -- from three manufacturing facilities spanning Minnesota, Idaho, and California.
APT's flagship product is the Advanced Cheese Vat (ACV), a patented design with three U.S. patents that improved yields, reduced maintenance, and set a new standard for process control in cheesemaking. Led by an on-staff Master Cheesemaker, APT provides recipe development, process optimization, and 24/7 engineering support.
We are not a software company that learned dairy. We are process engineers who built an AI system because we understood exactly what information our operators needed and couldn't get fast enough.
Get Started
Request a 30-minute live demo with a production ACV -- real questions, real data, real answers.
We'll be in touch within 24 hours to schedule your demo.