PDO Jabugo cellar curing with AI — governing the curve without replacing the master.
Curing in a PDO Jabugo cellar is the cellar master's call — but the temperature and humidity curve, hall by hall and vintage by vintage, leaves traces that a sensor at the right moment can put to work. iLEAN Edge captures those traces continuously and raises the alarm when the curve drifts. The master decides the adjustment — the data goes with him.
A Jabugo cellar is not a fridge — it is a curve, and a thermometer cannot read a curve.
Curing ham in a PDO Jabugo cellar rides on the Sierra climate: warm days to make the piece “sweat”, cold nights to settle it, shutter windows opened to the north or closed depending on the wind — and every vintage ends up with a fingerprint of its own. What decides final quality is not a fixed T/RH set point: it is the curve and its gradients over months.
- The classic cellar thermometer: one reading, once a day, written down by hand. Good enough to argue about, not to govern with.
- A generic datalogger: it records, but the data lives on an SD card that nobody ever cross-references with the vintage, the weight grade or the hall.
- The cellar master: has the judgement, but cannot be in four halls at once, or across two shifts.
- Differences between halls: the north hall and the south hall live different curves, and the master knows it — but the system does not, so every decision is made “from memory”.
When a vintage comes out odd and someone has to explain why, the data needed to reconstruct what happened is not there. And when the master retires, the fingerprint of every vintage leaves with him.
iLEAN Edge does not decide the curve — it captures it, hands it back and puts it at the master's disposal.
The cellar is the exact counter-example of the lights-out plant: here the cellar master's judgement is the asset. iLEAN is not here to automate curing; it is here to act as the putty between the master's knowledge and the plant's real data — so the judgement stays as a permanent capability and stops depending on one person's memory.
Edge records T/RH by hall, column and height, continuously. The agent compares it with the vintage's target curve. If it drifts, it raises the alarm. The master opens the shutter window — it does not open by itself.
The iLEAN pieces applied to curing in a PDO Jabugo cellar:
- iLEAN Edge (sensors) — discreet terminals per hall, geolocated by column and height. They record T/RH continuously, with enough resolution for the vertical gradient (floor vs. ceiling) and the difference between halls. They work with no network: if the cellar loses WiFi, Edge keeps recording and fires the local alert on the panel.
- iLEAN Connect — captures the master's knowledge: his voice when he explains why today he opens one extra shutter window, the photo of the panel in the modern hall that was stranded on an isolated screen, the vintage and the weight grade of the batch coming in. The vintage fingerprint is built from all of that, not from sensors alone.
- Agent — cross-references the real curve with the target curve, the history of past vintages and the weight grade of each batch. When the curve drifts, or when one hall falls outside the usual pattern for the vintage, it proposes an alert. The master decides. The agent only rides along.
Thermometer and notebook vs. a cellar governed with iLEAN Edge
| Aspect | Thermometer + notebook | With iLEAN Edge + Connect + Agent |
|---|---|---|
| Data granularity | 1 reading/day per hall | Continuous, by hall, column and height |
| Detecting curve drift | After the fact, during review | In the moment, with an alert to the master |
| Difference between halls | “The north hall runs colder” (by eye) | Quantified and comparable across vintages |
| Vintage fingerprint | The master's memory | Traced and reproducible |
| PDO dossier | Rebuilt by hand | Automatic, ready for audit |
| Capturing the master's judgement | Leaves with him when he retires | Stays as a capability of the cellar |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your cellar. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- A PDO Jabugo cellar with several halls (a mix of traditional and modern), tens of thousands of pieces curing over 24-36 months.
- Edge pilot in 1-2 halls (T/RH sensors + integration with the existing climate control where there is one + the master's voice via Connect). First value expected within a few weeks: the hall's fingerprint stops being an opinion and becomes data.
- Indicative payback between 5 and 9 months, depending on how often vintages have drifted in recent years and on the cost of a vintage that falls outside the standard.
- Hard levers: drifted-vintage risk down ≥ 30%, the master's fingerprint captured as an asset of the cellar, a PDO dossier ready for audit with nobody typing. Conservative figures, a defensible floor.
And the cellar master's reasonable doubt
“What if the algorithm decides to open a shutter window when I know it is not the moment?” — it does not decide. iLEAN captures, compares and alerts; the decisions that touch the cellar are signed by a person. Where models have been used for anchored tasks (reading a sensor, comparing it with a curve, flagging the deviation), the best ones brought error below 1.5%[1]. And even so, the three safety rings are there so that no critical operation runs without a human signature.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about AI-assisted curing in a PDO Jabugo cellar
Which variables define curing quality in a PDO Jabugo cellar?
In a Jabugo cellar the T/RH curve (temperature and relative humidity) is the criterion. It is not a fixed set point: it is a seasonal curve that exploits the climate of the Sierra and the “draft” through the shutter windows. What matters is each day's minimum and maximum temperature, sustained RH (with its peaks), the gradient between the floor and the ceiling of the hall, and the difference between the north hall and the south hall. The vintage does the rest: every batch leaves a fingerprint of its own.
How do you govern the T/RH curve by vintage and hall without replacing the master?
iLEAN Edge does not decide the curve — it captures it and hands it back. Discreet sensors in each hall record T/RH continuously, geolocated by column and height. The agent cross-references that fingerprint with the vintage, the weight grade and the target curing profile, and alerts the cellar master when the real curve drifts away from the agreed target curve. The master opens the shutter window — the AI does not open it for him.
Does iLEAN Edge work if the cellar has no stable connectivity?
Yes. Edge keeps working with no network as long as it has power: it records T/RH, fires local alerts and queues the trace to sync when the connection comes back. In a Sierra cellar with flaky WiFi, that is the difference between reliable data and a spreadsheet written after the fact. What is critical cannot depend on there being coverage.
How do you handle the difference between old and modern halls?
In a PDO Jabugo cellar, masonry halls with traditional shutter windows sit alongside more modern halls with climate control. iLEAN Edge adapts to each one: in the traditional hall it records T/RH and proposes when to open the draft; in the modern one it integrates with the existing climate control (without replacing it) and verifies that the real curve matches the requested one. The putty fills the gap between the two worlds — it does not force you to make them uniform.
Does the system let you reproduce the natural oxidation “pull” in every vintage?
It does not reproduce it — it documents and compares. The cellar master knows when the oxidation is going the right way. iLEAN Edge turns that knowledge into a vintage fingerprint of T/RH, gradients and window openings that can be compared with previous vintages. When the master retires, the fingerprint stays. And when a vintage strays from the pattern, the system says so — the master decides.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your cellar.
We work on the real data from your cellar, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See food industry