Ferment & Wash agent
Fermentation trajectory control for temperature, gravity and Brix, pH, and yeast health, tuned toward the target congener and ester profile before distillation begins.
What it does
Caskent models each wash as a live biochemical trajectory, then keeps the fermenter inside the temperature, sugar conversion, pH, and yeast-health envelope selected by the distiller.
- Forecasts gravity fall and heat generation across the run.
- Detects stalls, stress, nutrient drift, and contamination signals early.
- Shapes ester and congener precursors before they reach the still.
How it works
The agent fuses fermenter telemetry, lab checks, recipe intent, and historical runs. It recommends or applies bounded adjustments to jackets, cooling schedules, nutrients, and transfer timing.
- Shadow mode compares planned and actual trajectory.
- Assist mode proposes operator-reviewed actions.
- Graduated autonomy acts only inside approved limits.
Capabilities
Control features built for wash consistency and upstream quality.
Trajectory forecast
Predicts final gravity, completion time, thermal rise, and ester precursor range from live signals.
Thermal control
Tunes jacket schedules and ramp limits to protect yeast while keeping production on plan.
Health detection
Flags stuck fermentation, stress, nutrient deficit, and contamination patterns for review.
Recipe protection
Keeps site recipes, yeast programs, and sensory targets isolated to the tenant.
Transfer timing
Recommends when wash is ready for the still based on chemistry and operations constraints.
Inputs and signals
Every recommendation is grounded in signals your team can inspect.
| Signal | Source | What Caskent does |
|---|---|---|
| Temperature | Fermenter probes and jacket telemetry | Maintains heat generation inside the approved trajectory. |
| Gravity / Brix | Inline, sample, or lab readings | Forecasts attenuation and detects stalls before schedule risk grows. |
| pH | Probe or lab sample | Flags drift tied to yeast stress, contamination, or recipe mismatch. |
| Yeast health | Viability checks, nutrient data, historical runs | Recommends nutrient, cooling, or review actions inside guardrails. |
| Target profile | Recipe, standards, and sensory goals | Maps wash conditions to desired ester and congener precursors. |
Control sequence
A real ordered loop from model fit to operator-approved action.
- Step 1SENSE
Read the wash
Collect telemetry and lab context, then compare the live batch against recipe and historical envelopes.
- Step 2PREDICT
Forecast the endpoint
Predict completion time, pH path, gravity fall, and precursor profile before the deviation becomes costly.
- Step 3CONTROL
Act inside limits
Recommend or apply approved cooling, nutrient, and transfer adjustments, with the operator able to pause the loop.
Outcomes
Illustrative design-partner targets, measured against each site baseline.
Assurance note
Ferment & Wash runs under bounded control. Every recommendation and action is logged, grounded in available signals, and reviewable by the distiller. Human-in-the-loop checkpoints stay active across shadow, assist, and graduated autonomy modes.
Works with the rest of the stack
Fermentation quality compounds into cut control and downstream sensing.
- DCOpen →
Distill & Cut
Uses wash trajectory context to adapt the heads, hearts, and tails decision.
- SQOpen →
Spirit Quality
Correlates fermentation state with congeners, off-notes, clarity, and safety limits.
- DTOpen →
Digital Twin
Simulates recipe and run outcomes before production conditions are committed.
Fermentation stopped being a batch record we read after the fact. The team could see where the wash was headed and intervene before the still inherited the problem.
Control the wash before the cut.
Request access to evaluate Ferment & Wash on one fermenter, then connect the result to cut control and spirit-quality sensing.
Request access