Expertise
Manufacturing software, algorithmic design and machine learning
We build software from manufacturing records. With your engineers we handle and annotate the data, build models that support quoting, scheduling and quality decisions, and use generative design checked by algorithmic simulation. We also provide economic analysis for aerospace and defence.
1 Applied machine learning for manufacturers
Machine learning built on your production records
Most manufacturers hold fifteen years or more of quotes, drawings, tolerances and inspection results, and a team who know what they mean. We work with you to turn that history into software: we handle the data, annotate and structure it with your engineers, build and validate the models, and develop algorithmic engines that simulate outcomes and support decisions in the way your engineers make them.
Where the models run and who holds the data is agreed at the start. Your records are handled securely and confidentially throughout.
Records we can learn from
- QuotingQuotes, win and loss, actual cost and hours
- EngineeringDrawings, tolerances, revisions
- ProductionMachine logs, build parameters, scrap
- QualityInspection results, nonconformances, returns
- MaintenanceService history, fault codes
- StoresStock, lots, supplier lead times
How we work with you
- Data discovery and secure handlingWe audit, clean and protect what you share, under NDA, and agree where the data lives and who holds it.
- Annotation with your expertsYour engineers label examples with their judgement, so the system learns the reasoning as well as the numbers.
- Model and algorithm developmentWe build and train the models, and algorithmic engines that simulate price, lead time, tolerance risk and schedule.
- Validation against your own historyTested on jobs where you already know the answer, and measured against how the decision was made manually.
- Pilot, integrate and hand overRun alongside current practice on live work, then connected to quoting, scheduling and quality systems. People keep the final decision.
Software data inputs and outputs on the factory floor
Plan view of a production floor. Select a numbered point.
Quote intake
Reads
- CAD geometry and drawing callouts
- Material, quantity and tolerance class
- Past quotes: won, lost, actual cost and hours
Acts
- Drafts a price and lead time for estimator review
- Flags features that overran on similar past jobs
- Passes unusual requests to a person
What the software reads and does
- 1
Quote intake
Reads CAD, tolerances and past quotes
Does Drafts a price and lead time for the estimator to approve
- 2
Scheduling
Reads Machine queues, shifts and due dates
Does Sequences builds and re-plans when a machine stops
- 3
QC inspection
Reads Inspection results and build logs
Does Adjusts sampling and writes the inspection record
- 4
Maintenance
Reads Run hours, fault codes and service history
Does Books service windows in the quietest slot
- 5
Inventory
Reads Stock, lots and supplier lead times
Does Sets reorder points and keeps lot traceability
Data we learn from
Your records are secure and confidential.
2 Machine learning and AI in design
Generative design, checked by algorithmic simulation
Generative algorithms propose load paths and remove material that carries none, producing lightweight parts that additive manufacturing can build. Algorithmic simulation then repeats the analysis thousands of times with varied loads, dimensions and material properties, so a design is judged on the spread of outcomes rather than a single nominal case.
Typical use cases
| Part | Varied in the simulation run | Result |
|---|---|---|
| Brackets and mounts | Load direction, bolt preload, material scatter | Lighter geometry with material along the load paths |
| Brake and suspension parts | Pressure, friction, temperature | Branching structures sized to the peak load case |
| Fluid manifolds | Flow rate, wall thickness tolerance | Internal channels shaped for pressure drop |
| Lattice-filled housings | Cell size, relative density | Stiffness against mass for a printed shell |
- Loads and fixings
- Generate
- Simulate
- Validate
Stress field and load paths
For each load case, the colours show von Mises stress and the white lines show how load travels through the branches to the two bolts.
von Mises stress, MPa. Marker: example allowable, 440 MPa.
- Iteration
- 120 / 120
- Relative mass
- 0.63
- Max von Mises
- 342 MPa
- Mean, structure
- 140 MPa
- Reaction at A / B
- 33 / 67 %
Example simulation. Load runs from the four pad loads to the two bolts through the tree geometry shown.
3 Economic analysis
Policy and grant analysis for aerospace and defence investment
We work with local governments and agencies on economic investment policy for aerospace and defence, and on the grants that pay for it. The output is a report in the format below.
Source: ADS Group, Industry Facts and Figures, 9 July 2026.
Frequently asked questions
Do you build software for outside companies?
Yes. If a manual step costs you time on every job, we scope, build and test the software that removes it, trained on your own records and developed closely with your engineers.
Is customer data used to train models?
Files sent for a quote are secure and confidential, and are used to prepare that quote. Models for a factory-software engagement are trained on the data the customer supplies, and where they run and who holds the data is agreed at the start.
How much data do we need?
A data audit answers that. Fifteen years of quotes, tolerances and inspection results is a strong starting point, and a smaller archive works well for a single decision such as quoting. We handle the cleaning, annotation and model building with you.
Can you design a part with generative software and print it?
Yes. We set up the loads and fixings, generate and optimise the geometry, check it with algorithmic simulation across thousands of varied cases, and print it in a suitable material.
Do you advise local governments?
Yes. We work with local governments and agencies on economic investment policy for aerospace and defence, and on grant applications. See the economic analysis page.
Discuss a process to automate or a part to optimise
Tell us the decision that costs you time or the load case that drives the design.