A real-time smallware tracking & service optimization platform
for River Pointe Inn
At River Pointe Inn, the gap between a dish leaving the kitchen and the correct cutlery being present at the table is managed entirely by institutional memory, manual floor checks, and reactive scrambling at the polish station. There is no system. There is no signal. There is only experience — and when service accelerates, experience is not enough.
"The polish station should never be guessing. It should be executing a ranked work order generated by the tickets on the floor."
— Operational Design PrincipleThis proposal outlines a full-stack Cutlery Intelligence System — a purpose-built operational platform that connects every dish sold through Toast POS to a live demand calculation, monitors physical bin inventory through a sensor network, and generates a prioritized work order for the polish station in real time. The General Manager sees the entire floor. The polish staff sees exactly what to do next.
No system like this exists as a commercial product. This is built from scratch, owned entirely by River Pointe, and designed to scale.
Five layers work in concert. Every component is purpose-built, zero proprietary lock-in, and designed for a single-location fine dining operation.
Every sensor type — weight, RFID, infrared, camera, QR — posts to a single Cloudflare Worker endpoint. No sensor ever communicates with the database directly. The Worker normalizes all input, runs the demand calculation, and writes the result. This architecture means adding a new sensor type in the future is a matter of minutes, not weeks.
| Capability | Description | Phase |
|---|---|---|
| Menu-Cutlery Matrix | Every dish on the menu is mapped to its exact required pieces — dinner fork, fish knife, oyster fork — including tasting menu course sequences. | Phase 1 |
| Live Bin Inventory | Weight sensors and RFID readers report current piece count at every bin, every station, every 2 minutes. | Phase 1 |
| Demand Forecasting | Open Toast tickets feed a demand engine that calculates exactly which pieces are needed where in the next 15 and 30 minutes. | Phase 1 |
| Polish Priority Queue | A ranked, scored work order for the polish station — updated every 2 minutes — tells staff exactly what to polish and where to deliver it next. | Phase 1 |
| GM Floor Dashboard | All bins across all stations visible on one screen. Color-coded status, live queue, open ticket count, critical alerts. | Phase 1 |
| Staff Tablet View | Purpose-built mobile interface for polish station staff. Simple ranked list. One tap to mark complete. No training required. | Phase 1 |
| Full Event Audit Log | Every sensor reading, every scan event, every quantity change is logged immutably in Airtable with timestamp and source. | Phase 1 |
| Camera Computer Vision | Visual object counting at dish pit and polish station — verifies weight sensor readings and catches discrepancies. | Phase 2 |
| VoxIntel Integration | Passive RF monitoring correlated with staff and item movement patterns — surfaces behavioral insights invisible to sensors alone. | Phase 2 |
| Loss Detection | Statistical baseline per item type. Deviations beyond threshold trigger a GM alert — catches breakage and attrition before end-of-month inventory. | Phase 2 |
Each sensor type is deployed where it produces the highest signal. No location is instrumented for its own sake. Every sensor posts to the same endpoint.
One important hardware note: RFID tags specified for this system are rated for repeated high-temperature commercial dishwasher cycles. Epoxy-encased UHF Gen2 tags from manufacturers such as Xerafy and Confidex are designed for exactly this environment. Tags are applied to bins and racks — not individual pieces of cutlery — which dramatically reduces tag count and unit cost.
The polish queue is not a simple list of what is empty. It is a scored, ranked work order calculated from three independent signals — how short the bin is, how soon the pieces are needed, and how depleted the bin is relative to its par. Every item in the queue has a score. The highest score gets polished next.
The queue recalculates every two minutes via a Cloudflare Cron Trigger during service hours. As tickets open, close, and evolve — and as sensors report bin changes — the ranking shifts automatically. Staff at the polish station never need to assess urgency themselves. The system has already done it.
A table that waits for the correct fork is a table that remembers waiting. A table that never waits does not notice the system. That is the standard this platform enforces.
Software infrastructure costs are negligible. The majority of investment is hardware — sensors, tags, and mounting hardware. All software is self-hosted on infrastructure already in use.
| Item | Notes | Est. Cost |
|---|---|---|
| Load cell sensors (6 stations × 4 bins) | HX711 amplifier + ESP32 microcontroller per station | $280 – $420 |
| Industrial RFID tags (50 units) | Xerafy or equivalent; dishwasher-rated, on-metal | $180 – $300 |
| RFID readers (Bar + Storage) | UHF Gen2 USB or ethernet readers | $200 – $350 |
| IR break-beam sensors (3 thresholds) | Adafruit or equivalent; wired to microcontroller | $60 – $100 |
| Camera units (Phase 2) | Raspberry Pi HQ Camera + compute module | $180 – $280 |
| QR labels (printed) | Laminated, waterproof stock | $20 – $40 |
| Polish station tablet mount | Wall or rack mount; any Android/iPad tablet | $40 – $80 |
| Cloudflare Worker | Free tier — 100,000 requests/day included | $0 |
| Airtable | Free tier sufficient for Phase 1 | $0 |
| Cloudflare Pages (dashboard hosting) | Free tier | $0 |
| Total Hardware Investment | Phase 1 sensors + tags + mounts | $780 – $1,290 |
Phase 1 can be fully operational — software running, Toast integrated, QR scanning live — for zero hardware cost. Sensors are additive once the software foundation is proven in service.
This system was designed by and for River Pointe Inn. It is not a vendor product. It is not a subscription. It is operational infrastructure built to the specific rhythms of this kitchen, this floor, and this standard of service.
It is ready to build. The only question is when.