Platform — the full distillery autonomy stack

Caskent connects stills, fermenters, lab instruments, warehouses, bottling cells, and energy systems into one governed operations layer for spirits production.

One system of action from edge to cloud.

Edge runtime closes the plant loop. Cloud training improves policies from labeled runs, sensory feedback, quality outcomes, and rare-fault simulation.

PerceiveDecideActAudit
SYSTEM MAP
EDGEORCHESTRATORCONNECTORSCLOUDAUDIT LOG
Autonomy stack

Seven agents. One production loop.

Each agent links to a bounded control area and shares context through the site model.

  1. 01

    Ferment & Wash

    Trajectory control for temperature, pH, gravity, and yeast health.

    Explore
  2. 02

    Distill & Cut

    Real-time heads, hearts, and tails control.

    Explore
  3. 03

    Spirit Quality

    Congener, methanol, clarity, and off-note sensing.

    Explore
  4. 04

    Cask & Maturation

    Cask selection, extraction, oxidation, and angel's share.

    Explore
  5. 05

    Blend & Proof

    Blend recipes, proofing, dilution, and filtration to spec.

    Explore
  6. 06

    Robotic Bottling

    Fill, cork, label, inspect, and package with line feedback.

    Explore
  7. 07

    Energy & Yield

    Steam, heat recovery, yield, feints, and loss optimization.

    Explore

Closed perceive, decide, act control.

The platform reads plant signals, reasons against site limits, writes bounded setpoints, then records every action for review.

  • Plant telemetry and lab labels fused in one site state.
  • Policy router selects the safest approved model path.
  • Operators can pause, approve, or revert any action.
CONTROL LOOP
PERCEIVEDECIDEACTLOG

Prove the run in the digital twin.

Simulate cuts, congener profile, cask maturation, energy, and robot paths before committing policies to the plant.

DISTILLATION TWIN
SIM RUN 18
SPEC 82%
SIM RUN 44
SPEC 98%
SIM RUN 51
SPEC 91%
SELECT POLICY 44 → OPERATOR APPROVAL

Connectors and integrations.

Signals arrive at the edge. Approved actions write back only inside the operating envelope.

SystemSignalsControl write-back
Still telemetryABV, reflux, plate temperatures, flow, pressureReflux, heat, valve setpoints
FermenterTemperature, pH, gravity, Brix, yeast healthCooling, nutrient, agitation limits
Spectrometer / GCCongeners, methanol, volatile markersQuality gates and cut limits
Cask warehouseHumidity, fill, location, tasting, lossCask selection and movement plan
Bottling lineFill, cork, label, vision defectsRobot cell and rejection rules
Checkweigher / MESWeight, lot, batch, downtime, OEEBatch records and line gates

Architecture built for site reality.

Edge runtime and cloud training

On-site runtime keeps control close to the still, while the cloud improves models from approved records.

Model-agnostic router

Routes between frontier models, fine-tuned open models, and deterministic solvers by risk level.

pgvector, TimescaleDB, RAG

Time-series data, recipe knowledge, procedures, and safety standards stay queryable with citations.

REFERENCE ARCH
K3s EDGEROUTERRAGTRAINING

Assurance and trust.

Grounded outputs

Responses cite procedures, recipes, and safety rules used to support a recommendation.

Bounded autonomy

Hard safety and product-quality limits wrap every control policy.

Human checkpoints

Teams choose shadow, assist, or graduated autonomy by unit operation.

Audit trail

Operator approvals, model context, and write-backs are preserved for review.

Deployment

Connect, simulate, engage.

A staged rollout proves value before scaling across the site.

  1. Step 1

    Connect

    Attach Caskent edge to one still, fermenter, warehouse zone, or line.

    2 to 4 weeks
  2. Step 2

    Simulate

    Calibrate the twin against historical and live production runs.

    Twin fit
  3. Step 3

    Engage

    Move from shadow to assist to graduated autonomy as confidence grows.

    Governed

Outcomes tracked from day one.

Illustrative design-partner targets, measured against each site's baseline.

42%less off-spec spirit and feints
6.5%higher spirit yield per run
23%distillation energy recovered
100%actions logged for audit
FAQ

Platform questions.

Does Caskent replace existing PLCs and controls?

No. It sits above existing OT systems, reads validated signals, and writes approved setpoints through controlled interfaces.

Can we start with one workflow?

Yes. Most design partners start with a still, fermenter, bottling line, or quality workflow, then expand once value is proven.

What data is required?

Telemetry, lab results, sensory notes, batch records, and procedures are enough to begin. More labeled outcomes improve the policies.

How is craft preserved?

The master distiller sets the spec, limits, recipes, and approval mode. The platform executes within that envelope.

Request access to the autonomy stack.

Route one still or line through Caskent and measure quality, yield, energy, or assurance impact.

We reply within two business days.