One loop, from the first experiment to the running plant.

Self-driving lab: Experiments that choose the next experiment

A robot prepares and runs reactions around the clock. After every result, DecaOS updates its model of the chemistry and picks the experiment that will teach it most, instead of working through a fixed grid.

Scale-up: Pilot crystalliser

Crystals that form nicely in a flask can turn fine and slow to filter at scale. DecaOS predicts the cooling curve for 50 litres from the lab data, and the pilot proves it.

Crystalliser: Crystals of the right size, every batch

The active ingredient is crystallised by cooling. Cool too fast and fine crystals clog the filter; too slow and the batch takes hours longer. DecaOS reads the in-line particle probe and reshapes the cooling curve while the batch runs.

Filling line: Right first time, line after line

Vials are filled, stoppered and inspected at speed. DecaOS connects lab, process and line data, so a drift in fill weight is caught and corrected before a batch is lost.

The DecaOS loop

Every batch record from the crystalliser and the filling line flows back to DecaOS, so the next molecule's development starts from what this plant actually did.

One loop from the first experiment to the last vial

• Development in half the time. The self-driving lab learns which experiments matter.
• Scale-up with fewer surprises. The model built in the lab carries into the plant.
• Lines that hold their quality. Deviations are caught as they start, not after release testing.

Self-driving lab: Cooking and bleaching recipes, tested by robots

Wood, water and chemicals change from season to season. A self-driving lab cooks and bleaches small pulp samples around the clock, and DecaOS learns how each recipe changes yield, brightness and strength before the mill tries it.

Scale-up: Pilot digester

A small digester cooks real chips with the recipes the lab found best. DecaOS predicts the kappa number and the yield before each cook, and learns from the gap.

Digester: Cooking wood into pulp, on target

Wood chips are cooked with chemicals and steam until the lignin that binds the fibres dissolves. DecaOS predicts the kappa number before the cook ends and trims temperature and time, so the pulp leaves on target with less steam.

Bleaching: Brighter pulp with fewer chemicals

The brown pulp is bleached in stages. Overdosing wastes chemicals and weakens the fibres; underdosing misses brightness. DecaOS doses each stage from what the stage before it actually did.

Dryer section: Dry enough, not too dry

The wet sheet runs over dozens of steam-heated cylinders. Most mills over-dry to stay safe, which burns steam and makes paper brittle. DecaOS holds the moisture at the reel right on target and spots a sheet break before it happens.

The DecaOS loop

Readings from the digester, the bleach plant and the dryer flow back to DecaOS, so the lab tests next season's wood and chemicals against what the mill actually did.

Less steam, every reel on spec

• Less steam and power across pulping, bleaching and drying.
• Every reel on spec. Moisture, brightness and strength are held, not chased.
• Fewer breaks and recooks. Problems are seen minutes before they start.

Self-driving lab: Catalysts and cracking conditions, tested around the clock

Small reactors test feeds, catalysts and cracking conditions in parallel. DecaOS chooses each next test from everything measured so far, so the best conditions are found in weeks, not years.

Scale-up: Pilot cracker

One tube heated like a full furnace coil shows how a new feed or catalyst cracks and cokes. DecaOS compares it with its prediction before any furnace changes.

Cracking furnace: Hot enough to crack, not to coke

Feed is cracked into ethylene inside tubes running through a furnace at over 800 °C. Run hotter and the yield rises, but coke builds up faster and the furnace must stop for cleaning. DecaOS balances both, furnace by furnace, hour by hour.

Distillation column: Close to flooding, never over

Columns separate the products. Each tray holds liquid while vapour bubbles up through it; push too much vapour and the froth fills the gap between trays and the column floods. Most columns run far from that limit to be safe. DecaOS runs them closer, safely.

Flare and emissions: Less flaring, fewer tonnes of CO₂

Upsets send gas to the flare, where it is burnt and lost. By keeping furnaces and columns steady, DecaOS prevents most upsets before they start, and tracks the CO₂ of every unit.

The DecaOS loop

Furnace and column data flow back to DecaOS, so every new feed and catalyst is tested in the lab against what the furnaces actually did.

Closer to the optimum, around the clock

• Less fuel and CO₂ from furnaces running at their best balance of yield and run length.
• More product from the same columns, run closer to their limits, safely.
• Fewer upsets and less flaring, because drift is corrected before it grows.

Self-driving lab: Finding better catalysts, faster

New catalysts and electrolytes decide how cheap green hydrogen and clean fuels can be. A self-driving lab mixes, tests and measures candidates around the clock, and DecaOS chooses each next recipe from everything measured so far.

Scale-up: Test stack

A short stack runs the best materials from the lab on a real wind-and-solar power profile. DecaOS predicts its efficiency and wear, and learns from the gap before the plant is built.

Electrolyser: More hydrogen per megawatt-hour

Water is split into hydrogen and oxygen in stacks of cells. Efficiency drifts with temperature, load and age. As wind and solar power rise and fall, DecaOS sets each stack to make the most hydrogen from the power available, and slows its wear.

Carbon capture: Capture more CO₂ with less steam

Flue gas rises through an absorber while a solvent soaks up its CO₂; steam then strips the CO₂ back out. Steam is the biggest cost. DecaOS matches the solvent flow and the steam to the gas coming in, minute by minute.

The DecaOS loop

Stack and capture data flow back to DecaOS, so the lab's next materials are tested against how the plant's stacks actually age.

Cleaner molecules, sooner and cheaper

• Better materials, found faster, by a lab that learns from every test.
• More hydrogen from the same power, with stacks that last longer.
• More CO₂ captured for less steam, minute by minute.

Self-driving lab: Strains and recipes, tested in mini-fermenters

Dozens of small fermenters run strains, feeds and temperatures side by side. DecaOS reads each one as it grows and picks the next recipe, so a better product and a faster process are found in weeks.

Scale-up: Pilot fermenter

The best recipe from the mini-fermenters runs at 200 litres, where mixing and cooling are slower. DecaOS predicts how the batch will behave and adjusts it as it runs.

Fermentation: Every fermentation like the best one

Yeast turns sugar into alcohol, flavour and CO₂, and no two batches behave exactly alike. DecaOS learns each batch's pace from its first hours and adjusts temperature and feeding, so batches finish on time and on taste.

Cooking and pasteurising: Safe, without overcooking

Products are heated to a set temperature for a set time to make them safe. Most lines add a wide margin, which costs energy and taste. DecaOS holds the real temperature in the product on target, with the safety margin proven, not guessed.

Spray dryer: Powder at the right moisture

Milk, coffee and other liquids are sprayed into hot air and dry into powder in seconds. Too wet and it cakes, too dry and energy and yield are lost. DecaOS steers the hot air to the moisture you need.

The DecaOS loop

Every batch's fermentation, cooking and drying data flows back to DecaOS, so new recipes start from how this plant's batches really behave.

Quality on spec, with less energy and water

• Every batch like the best one, from fermentation to drying.
• Less steam, power and water, because margins are proven, not guessed.
• Less waste: fewer batches off spec, less product lost in the dryer.

Self-driving lab: Better routes to new molecules

There are often dozens of ways to make a molecule. The self-driving lab tries routes, solvents and catalysts in parallel, and DecaOS steers the search towards the cheapest, cleanest route that scales.

Scale-up: Pilot reactor

The best route from the lab runs at 100 litres, where heat builds up faster than it escapes. DecaOS predicts the safe dosing rate and the yield, and the pilot proves both.

Batch reactor: Reactors that learn from every batch

A batch reactor follows a recipe: add, heat, hold, cool. But raw materials and equipment vary. DecaOS watches each batch as it runs and adjusts dosing and temperature, so every batch ends at its best.

Purification: Purity without the extra pass

Crude product is purified by distillation. When in doubt, plants run a second pass and burn the energy twice. DecaOS predicts the purity from live data, so one pass is enough.

The DecaOS loop

Every batch record flows back to DecaOS, so the next molecule's route is chosen with what this plant's reactors actually did.

Better routes, better batches

• Cheaper, cleaner routes, found in weeks by a lab that learns.
• Every batch at its best, with reactors that adjust as they run.
• Release sooner, with purity predicted, not re-tested.

Self-driving lab: Dye recipes matched by robots

A robot lab dyes small swatches and measures their colour. DecaOS learns how fibre, dye, salt and temperature set the shade, so each new colour is matched in a few tries and comes out right first time in the mill.

Scale-up: Sample dyeing machine

The lab's recipe dyes 10 kilograms of real fabric in a small jet machine. DecaOS predicts the shade before the run ends and corrects it, so the mill gets a proven recipe.

Dyeing machine: The right shade, first time

Fabric circulates in hot dye liquor until it takes up the colour. When a batch misses the shade it must be stripped and dyed again, using the water, heat and chemicals twice. DecaOS predicts the shade during the run and corrects it in time.

Stenter: Dry and set, without wasting heat

The stenter stretches the fabric to width and dries it in a long hot-air oven. Ovens often run hotter and slower than needed. DecaOS sets the temperature, airflow and speed for each fabric, so it leaves dry and on width with less gas.

Water recovery: Reuse the water, recover the heat

Hot wastewater from dyeing carries heat and chemicals. DecaOS plans which batches can reuse which water and recovers the heat, so less fresh water and steam come in.

The DecaOS loop

Every dye lot and stenter run flows back to DecaOS, so each new colour is matched in the lab with what the mill's machines actually do.

Right first time, with less water and heat

• The right shade, first time, so no batch is dyed twice.
• Less gas in finishing, with each fabric dried just enough.
• Water and heat reused, batch after batch.