Every chip, every lithography machine, every aircraft and every weapon system is built from parts that someone learned to make by hand. Those parts come from shops of ten to two hundred and fifty people.
An estimator determines what a part requires. A programmer or work-preparation lead translates that into operations, setups, tooling and machine choices. Both build years of unique knowledge and experience, but that knowledge is rarely captured in a structured way.
ERP systems record orders, materials, hours and cost. CAD describes the design. CAM translates operations into machine programs.
What sits between those systems is far less well captured: why an experienced machinist chooses a given operation, which setup makes sense, which machine fits, and what a tolerance costs in time and money. That is the knowledge layer Subduxion models.
Four foundations
Learning from real production, part requirements linked to the right geometry, working on top of the systems a shop already uses, and confidentiality: four things that have to hold before a calculation is traceable to its inputs and assumptions. We publish each one with its source.
Learning from real production
A calculation predicts what an operation will require. Production shows what actually happened. Your machine already logs its own cycle time, over MTConnect and OPC UA among others. Since 12 September 2025 the EU Data Act gives you the right to that data in machine-readable form; from 12 September 2026 new equipment must offer that access by design. By putting planned and realised time, machines, materials and operations side by side, the model can learn what actually works in a specific shop. We do not present that feedback loop as running everywhere today.
Requirements linked to the right geometry
A tolerance, fit or surface finish only means something once it is clear which face, hole or other geometric element it applies to. Blake is learning to connect information from 2D drawings to the 3D model. We do not claim to work directly from a full model-based definition and the PMI it carries.
Built on top of the systems a shop already uses
Blake is designed to work alongside the ERP, CAD and CAM systems a shop already uses, not to replace them. The model brings together information from those systems around a single part and the decisions needed to make it. Start from an upload of a 2D drawing or 3D model. Other intake channels are not available today.
Built for sensitive manufacturing data
Technical drawings, 3D models and production data often carry confidential knowledge from both supplier and end customer. That is why we apply high security standards to storage, processing and access, and develop AI with clear boundaries around confidentiality, control and responsible use. Where needed, processing can be set up within the EU or within the customer's own environment.
What a manufacturing model has to understand
Geometry determines what has to be removed. It does not by itself determine what a part will cost, how long it will take, or whether a tolerance will hold in production.
More than cycle time
Historical cycle time alone is not enough. A model that has to understand production needs the part, the operations, the machines and the measured outcomes, and above all the connection between them. That connection is what makes it possible to go beyond predicting time.
Physical process behaviour
In machining, outcomes are shaped by how material, cutting tool, machine dynamics and process parameters interact: through cutting forces, heat, deflection, vibration and tool wear. Our research investigates how knowledge of that physical behaviour can complement patterns learned from production data.
Every process is different
Milling, turning and EDM are not the same problem. Cutting is bounded by chip formation and machine dynamics, grinding by how energy turns into heat in the contact zone, EDM by electro-thermal erosion. Each has its own constraints and its own failure modes, so each needs its own predictive treatment. What can be shared is the description of the part.
One shared representation of geometry, requirements and factory. A model per process. Each line shows its basis, for review.
Research program
Represent
Build a shared representation of the part, its requirements and the factory context in which it has to be made.
Predict
Model how manufacturing processes and production choices affect manufacturability, route, machine fit, time and cost.
Learn
Connect predictions to measured production outcomes, so evidence from the physical process can improve future reasoning.
The long-term research direction is a world model that can compare possible ways of making a part before production starts.
Manufacturing understanding benchmark
We evaluate how much manufacturing intent a system can recover as a part moves from geometry-led to increasingly requirement-dependent complexity. The benchmark separates recognising geometry from understanding what that geometry means for manufacturing. The same evaluation principles apply at every level, and results are published only where they are repeatable and traceable to their source.
Geometry-led
Shape carries most of the manufacturing intent: a plate with three through-holes, general tolerances, one material.
Local requirements
Fits, tolerances and surface requirements alter how individual features have to be manufactured: a turned-milled part with two fits and a roughness requirement on one face.
Referenced requirements
Datums and geometric tolerances connect multiple geometric entities: a position tolerance on a hole pattern relative to one or two datums, with a separate heat treatment.
Interdependent definition
Manufacturing intent depends on interacting requirements: a full datum reference frame, an assembly with a bill-of-materials line, or a sheet-metal part with bend-order-dependent tolerances.
System-level reasoning
Decisions depend on multiple interacting requirements and on the consequences of process choices: composite tolerancing, a simultaneous requirement across features, a virtual-condition calculation that affects the fit.
The axis is fixed in advance.
The definition of complexity is set before a test runs, not after.
Complexity comes from the input, never from our own output.
A part's complexity is set by the generator, not by what our own system makes of it.
We generate the parts ourselves.
The ground truth is fixed in advance, and no real customer part ever goes through it.
Traceable to the source.
A result is only published when it is repeatable and can be traced back to the input it came from.
Confidence bands.
Every result shows a range, never a single point alone.