I connect the plant floor to the software that runs it.
I design and build the data layer between machines, PLCs and SCADA and the business systems above them: pipelines, historians, dashboards and custom apps. Fewer spreadsheets, more decisions made on real data.
Data acquisition and contextualization
I connect PLCs, SCADA and historians over OPC UA, MQTT or Modbus and turn loose signals into a data model with context: line, shift, order, product.
Platforms and dashboards
Time-series stores, reliable pipelines and OEE, quality and energy dashboards plant people actually use. Alerts that matter, not noise.
Custom applications
Full-stack apps for floor and office: traceability, maintenance, automatic reporting and ERP/MES integrations.
Cases
One historian for six plants
Every plant reported on its own spreadsheet with its own criteria. I unified acquisition over MQTT Sparkplug B into a time-series store with a shared line and shift model.
Real-time OEE without touching the PLCs
Passive OPC UA reads and availability, performance and quality computed at the edge. Supervisors stopped estimating downtime from memory.
Anomaly detection on consumption
Per-asset electrical consumption pipeline with a per-shift baseline and deviation alerts. It surfaced phantom loads nobody was watching.
End-to-end batch traceability
A web app tying raw material, process parameters and lab results to every batch, with an auditable log and signable reports.
MES ↔ ERP integration
Replaced manual re-keying between floor and ERP with a queued integration with retries and daily reconciliation.
Predictive maintenance, the honest version
Before the model, the data: instrumentation, cleanup and failure history. Only then a simple vibration model with thresholds reviewed by the team.
Blog
Your problem is not AI — nobody trusts the data
Before talking predictive models, there are three questions almost no plant can answer about its own data.
MQTT Sparkplug B: why I stopped writing point-to-point integrations
Six plants, four protocols and a separate integration per pair. The predictable ending, and how I reversed it.
How we went from a 4-hour report to 8 minutes
The batch report was assembled by a person copying from three systems. It did not need a year-long project.
Five signs nobody uses your dashboard
It looks spectacular at the kickoff and is dead three months later. The symptoms are always the same.
One foot on the plant floor, one in the repo
I work on the OT/IT border: where a value is born in a sensor and has to end up on a dashboard someone reads at 6am. I speak both languages — industrial protocols and modern software architecture — and that crossing is where I add the most value.
End to end: floor assessment, architecture, implementation, commissioning and training the internal team. No eternal dependency: I hand over documented and transferable systems.
- Remote — LATAM / EU
- Project or retainer
- Spanish / English
- Documented and transferable
Got data trapped on the plant floor?
Describe the problem in two paragraphs. You get an honest diagnosis back, and if it is a good fit, a concrete proposal.