AsphoriX
An engineer inspecting a gold Webb telescope mirror with a light.

Technology

Better measurement. Smarter correction.

We’re developing software that turns multiple measurements into a precise polishing plan—and uses each run’s results to make future corrections more predictable.

Webb mirror inspection. Industry context: NASA/Chris Gunn.

01 / The manufacturing loop

Precision, with fewer corrective passes.

Measure the surface, bring the data together, and calculate the next polishing pass. Then measure again. The loop continues until the mirror meets its specification.

Surface formWavefrontLocal texture
Surface form
Wavefront
Local texture
Local texture samples
Target surfaceTool removal model
LessMore
Precision metrologyConsolidated surface datasetCorrection planningCNC execution

Precision metrology

See the surface clearly.

Capture overall shape, wavefront error, and fine surface texture. Different instruments reveal different parts of the picture.

Fusion metrology

Many inputs. One surface.

Align compatible measurements in a common coordinate system. Combine their strengths and keep local texture samples linked to their measured locations.

Correction planning

Plan a precise correction.

Compare the measured surface with the target. Use calibrated tool behavior to calculate dwell time, toolpath, and process conditions.

CNC execution

Put the plan to work.

Execute the planned pass and record process conditions. Then measure the result to check precision and guide any further correction.

Repeat until the surface meets specification.

02 / The learning loop

Improve the mirror. Then improve the prediction.

The manufacturing loop corrects the surface. Alongside it, accumulated measurements and process history help refine the models behind future correction plans—across passes and repeat parts.

Compare with reality

Predicted
Measured

Planned vs actual conditions

Compare what the plan predicted with what the next measurement reveals. Include how actual process conditions differed from the plan.

Build process memory

Surface measurements
Tool and process conditions
Predicted and measured outcomes

Connect each starting surface, polishing plan, and measured outcome. Preserve the material, tool, and process context behind every run.

Refine the model

AI learningPredictedMeasuredCorrection planner

Use accumulated runs to train and refine removal predictions. Feed those improvements into dwell time, toolpath, and future process conditions.

Every run refines the model.

Refined models feed back into the planner, helping reduce rework across passes and repeat parts.

What feeds the learning loop?
Surface vs target
The remaining deviation tells the planner what the mirror still needs.
Predicted vs measured result
The prediction error shows where the removal model needs to improve.
Planned vs actual conditions
Differences in pressure, speed, tool wear, and other conditions help explain the outcome.

03 / The development path

From one cell to connected production.

Start with one integrated cell. Repeat its process in a production line, then connect multiple lines through shared process and quality data.

Cell → line → factory

One connected cell

Bring polishing, measurement, and correction software together as one repeatable manufacturing process.

  • Precision polishing and metrology
  • Connected correction software
  • Part handling and process data
Cell → line → factory

A repeatable line

Replicate the cell, connect material flow, and carry process knowledge forward as capacity grows.

  • Repeated manufacturing cells
  • Coordinated part flow
  • Shared recipes and quality history
Cell → line → factory

A connected factory

Bring multiple production lines onto a common data system, connecting equipment, quality, and capacity across the factory.

  • Multiple connected lines
  • Factory infrastructure and handling
  • One process and quality data system