Loading experience
Telemetry that makes structures talk.
With CMS, wireless sensors read what wood, concrete and structures are doing, and the platform we built turns that stream into live dashboards and early alerts.
Temperature, vibration & humidity in real time
Telemetry that makes structures talk.
Wireless sensors measure what the wood, the concrete and the structure are doing. The platform we built turns that stream into live dashboards, early alerts and engineering decisions.
Continuous readings from every sensor on site
Deliverables: platform, design system, data model, hardware
Time-series model for every sensor type
At a glance
- Client
- CMS (Construction Monitoring System)
- Sector
- Structural monitoring (construction · IoT)
- Projects
- UPM, WTC Free Zone (with Saceem, Teyma and Stiler)
- Scope
- Telemetry platform, design system, data model and field-hardware integration
- Status
- Delivered and in use (AI-on-data modules next in the pipeline)
The starting point.
CMS monitors real structures for Uruguay’s biggest builders. The field hardware was solid: thermocouples, accelerometers, inclinometers and LoRa gateways, all working. The software layer wasn’t keeping up. That was our part.
- Thermocouples
- Accelerometers
- Inclinometers
- Weather stations
- LoRa gateways
Four deliverables.
Telemetry Platform
Ingests sensor data in real time and turns it into live dashboards, alerts and reports.
Design System
One visual language across every screen, built for engineers reading data in the field.
Data Structure
A time-series model that scales from one thermocouple to an entire monitored structure.
Hardware Integration
A direct pipeline from LoRa gateways on site to the cloud, with no manual steps.
The structure’s vital signs, live.
Concrete curing at UPM, a footbridge vibrating under pedestrians, a wall slowly tilting. Each sensor reports continuously and the platform processes every reading as it arrives: live visualisation, configurable thresholds, and alerts that fire before an incident, not after.
A heart-rate monitor for buildings: continuous vitals, and an alarm the moment something drifts out of range.
Engineering data, readable on a job site.
The people reading these screens are field engineers, not analysts at a desk. We built the design system from scratch, with components, colour logic for alert states and data-dense layouts, so every chart reads the same way on a laptop in the office and on a phone in full sunlight next to the pour.
One model, every sensor.
Under the dashboards is the real asset, a time-series structure that maps how monitoring actually works across projects, structures, devices, channels and readings. New sensor types plug into the same model, historical data stays queryable, and the whole thing is analytics-ready.
From the field to the cloud.
Data is only real-time if the pipeline is. We integrated CMS’s field hardware straight into the ingestion layer, so a reading taken inside a concrete mass becomes a point on a dashboard without a single manual step.
Built AI-native.
The entire platform, from application code and design system to data model and tests, was engineered with AI coding agents, directed and reviewed by our senior team. Here AI was the workforce that built the product.
- Months of hand-written code
- Design system treated as a luxury
- Tests written “when there’s time”
- Big team, big budget
- Agents generate; seniors direct, review and own every line
- Components and docs generated alongside the product
- Test coverage produced with the code, not after it
- Small senior team, startup cost, enterprise output
What’s next.
The data structure was built for this. On the roadmap:
- Anomaly detection on live sensor streams, with alerts driven by patterns in the data
- Predictive warnings: flag a curve going wrong hours before the limit is crossed
- Engineering reports generated straight from raw telemetry
- Natural-language queries that ask the structure how it’s doing