Menu

Plant models built on your data. No platform to buy.

We build first-principles models, soft sensors, forecasts and optimisers around your plant, from one defined problem to a full-scale project your engineers can check.

Start with a $500 data audit See the evidence
Process flow illustration with a soft sensor Temperature and flow signals from a process line feed a calculated quality estimate, shown as a continuous trend between occasional lab samples. TI FI QY Exchanger Separator Column Sample point Calculated estimate Lab sample
Illustration, not plant data.

What are you hearing at your plant?

Soft-sensor proof of concept

Estimate a lab-measured quality between samples, using the signals you already record.

You get
A backtest on held-out data showing whether the estimate beats your current practice.
Price
From $2,000
See scope and deliverables

Sensor and mass-balance check

Find the meters and analysers that disagree with each other and with basic conservation.

You get
A ranked list of suspect instruments and the balance gaps behind each one.
Price
From $1,000
See scope and deliverables

Data and KPI audit

Map what you record, test whether it is fit for use, and tie it to the decisions that matter.

You get
A data-quality report and a short list of the projects your data can support.
Price
From $500
See scope and deliverables

Fouling or performance prediction

Forecast duty loss on exchangers and similar equipment from existing historian data.

You get
A forecast tested on past campaigns, with its misses and limits documented.
Price
From $2,000
See scope and deliverables

Surrogate or reduced-order model

Build a fast surrogate or reduced-order model of your simulation, validated against first principles.

You get
A fast model with its valid range, its error against the full model and its balance checks.
Price
From $3,000
See scope and deliverables

First-principles process models

Build process models from mass and energy balances, thermodynamics and kinetics, from one unit to a full-scale project.

You get
A process model scoped to your need, calibrated and tested against your plant data.
Price
Custom quote
See scope and deliverables

Real-time optimisation

Use live plant data and a validated model to recommend better operating points as conditions change.

You get
Advisory recommendations, tested in a shadow run before anyone relies on them.
Price
Custom quote
See scope and deliverables

Energy optimisation

Find where energy is lost across furnaces, heat recovery, steam and compression, and rank what to change.

You get
A ranked list of energy opportunities, calculated from your plant data with a model behind each.
Price
Custom quote
See scope and deliverables

For consultancies

White-label subcontract

Modelling capacity for engineering consultancies, delivered under your name.

You get
Reviewable models and reports your team can present as its own.
Price
From $50/hour
See scope and deliverables

The capability of a platform, scoped to what you need.

Typical platformWorking with Slipstream
Licence and long rolloutWork scoped to what you need, from one problem to a full-scale build
Book a demo to learn anythingScope and deliverables on the page, with prices for standard work
Headline ROI percentagesTested against simple baselines and first principles, including where the model loses
You rent their productYou keep the model, the code and the report

Five stages. You can stop at any gate.

Each stage ends with a decision, so you never pay for work the evidence doesn’t support.

  1. Stage 1

    Audit

    Map your tags, check units and timestamps, and test whether the data can answer the question.

    Gate Is the data fit for use?

  2. Stage 2

    Baseline

    Agree success criteria up front, freeze a test period and score simple methods first.

    Gate Is there room to beat the baseline?

  3. Stage 3

    Model

    Fit the model and test it on data it has never seen, including its misses.

    Gate Does it beat the baseline?

  4. Stage 4

    Shadow run

    Run read-only forecasts alongside operations, with your engineers reviewing each one.

    Gate Is it useful in practice?

  5. Stage 5

    Handover

    Hand over the model, code, documentation and the conditions where it should not be trusted.

    Gate Who owns it from here?

Our hybrid model lost this test. We publish that too.

We built an exchanger fouling benchmark on 88 synthetic campaigns. On campaigns that shifted away from the training conditions, a simple trend predicted threshold timing better than our grey hybrid: 1.38 days of error against 2.25.

The grey model’s uncertainty band also averaged 44.65 K, too wide to plan a cleaning around. This is why every project scores simple methods first.

Read the fouling case study
Simulated data

Error in predicting when a fouling limit is reached

Mean timing error in days on 24 shifted synthetic test campaigns. Lower is better.

  1. Simple trend 1.38 Simple trend Timing error 1.38 days Temperature error 1.70 K Mean band width 8.86 K
  2. Grey hybrid 2.25 Grey hybrid Timing error 2.25 days Temperature error 1.72 K Mean band width 44.65 K
  3. White hybrid 3.17 White hybrid Timing error 3.17 days Temperature error 1.53 K Mean band width 8.26 K
  4. Black hybrid 3.23 Black hybrid Timing error 3.23 days Temperature error 2.05 K Mean band width 7.99 K
  5. Fixed rate 5.85 Fixed rate Timing error 5.85 days Temperature error 2.59 K Mean band width 10.72 K
  6. Direct ML 15.15 Direct ML Timing error 15.15 days Temperature error 5.87 K Mean band width 9.78 K
View as table
Method Timing error (days) Temperature RMSE (K) Mean band width (K) Band coverage
Simple trend 1.38 1.70 8.86 96%
Grey hybrid 2.25 1.72 44.65 100%
White hybrid 3.17 1.53 8.26 98%
Black hybrid 3.23 2.05 7.99 78%
Fixed rate 5.85 2.59 10.72 80%
Direct ML 15.15 5.87 9.78 19%

5 model seeds on one synthetic fleet. Persistence is left out because it never predicted a crossing, so it has no timing error.

AI that has to obey the physics.

Every data-driven model, surrogate and ROM we deliver is validated against first principles before you rely on it.

Conservation
Outputs must close mass and energy balances within an agreed tolerance.
Physical limits
Predictions must respect bounds such as temperatures, compositions and efficiencies that are physically possible.
Correct response
When an input moves, the output has to move the way the physics says it should.
Full-model comparison
Surrogates and ROMs are tested against the first-principles model on runs they never saw.
Known range
Every model records where it was tested, and flags cases outside that range.

Only oil and gas, from wellhead to refinery.

Upstream

Wells, production facilities and produced water

  • Estimate rates between well tests
  • Spot gradual pump and compressor decline
  • Reconcile allocation against metered volumes
  • Recommend better operating points

Midstream, gas and LNG

Pipelines, gas processing and liquefaction

  • Estimate gas quality between lab samples
  • Track exchanger and compressor performance
  • Close custody and inventory balances
  • Cut fuel gas and compression energy

Refining

Process units, preheat trains and utilities

  • Forecast crude preheat fouling
  • Estimate product quality between samples
  • Speed up slow process simulations
  • Reduce furnace and steam energy use

Start small. Know the price first.

Starting prices for a typical scope. Full-scale and optimisation work is quoted after a short scoping call.

Your plant data is treated as confidential.

NDA first
We sign before any data moves.
Read-only
We never need write access to a control system.
Your data stays yours
Nothing is published or reused without your written agreement.
Engineers decide
Models inform decisions. They don’t act on the plant.

What engineers usually ask.

Can you build a full process model from scratch?

Yes. We build first-principles models to the size you need, from a single unit to a full-scale project.

How do you know an AI model is physically sound?

We check it against first principles: conservation of mass and energy, physical limits and the direction of its response. Surrogates and ROMs are also compared with the full model.

Does real-time optimisation change our setpoints?

No. It recommends operating points from a read-only feed, and your engineers decide what to change.

Do we need to buy any software?

No. You receive the model, code and documentation. There is no platform licence.

How long does a project take?

An audit is short. A forecasting study on one to three assets is typically planned at four to eight weeks, depending on the state of the data.

What if our data isn’t good enough?

That is what the audit is for. If the data can’t support the work, we say so and recommend a narrower study or stop there.

What if a simple method beats the model?

Then that is the result, and we’ll tell you. A simple method you can trust beats a complex one you can’t.

Do you work outside oil and gas?

No. We focus on upstream, midstream, gas processing, LNG and refining.

Tell us what you’re hearing at your plant.

Start a project