Clinton Williams / Massachusetts

I built Atlas to put past experiments back to work.

A finished production bag was headed for scrap. Atlas found the tests that could tell Quality whether the seal was acceptable and drafted a cited assessment. Quality reviewed it and kept the bag. I conceived the agent, built it solo and market it to the teams who use it.

ATLAS / PRODUCTION CASEClinton Williams

One finished bag.

Deviation
The seal ran outside its recorded setting.
Atlas retrieved
Prior seal experiments, validation reports and product-quality requirements.
Quality decided
The evidence applied. This assembly was accepted in one working session.
Anonymized case record · read the evidence

From a cleanroom shift to engineering software.

I’m a mechanical and robotics engineer. I supervised 24 cleanroom operators, then led the team that raised finished mixer-bag output from 800 to 2,000 a month. When the shift board kept losing the history of stops, I wrote the operator software and its database. When a supplier blamed cooling for a defect on the same edge of several bags, I asked to see the plate inspection. The heater face needed rework.

Atlas grew out of a similar problem in R&D. An engineer can find a familiar test and still apply it to the wrong assembly. I built the search, the evidence checks and the draft that a person can challenge. The 15-engineer alpha group logged 1,689 model calls in an 18-day July observation. One question can require several calls. The production record documents one accepted bag.

Selected work

Decisions from
the factory floor.

These projects show when I chose to write software, measure the material directly, or change how people worked around a machine.

Magnetic board showing machine assignments and production status.Before · board
My operator interface for machine state, production work and service requests.After · operator screen
Original shift board and the interface I built.
02

Digital production

I replaced the erasable shift board with a record operators and maintenance could use the next morning.

Trace a changeover
Four thermocouple traces in a single bond cycle over time. The channel peaks range from 228.9 to 244.9 °F.
Original thermocouple data · one bonding cycle.
03

Process validation

All 33 pressure-tested bags passed. The test sheets omitted the settings used to make them. That distinction changed what I could sign off.

Read the process decision
04 / COVID manufacturing ramp

800 to 2,000 finished bags a month.

Finished mixer bags per month
See the ramp

The advice was to skip the finisher. What happened next?

Genysys is my independent venture. Its Arciv prototype keeps the training advice and the athlete’s choice in separate records. In this illustrated episode the athlete does the optional finisher. The code and screens exist. The first commercial test is whether a team will use the record, find it helpful and pay to continue.

Original Arciv prototype screens illustrating the recommendation and the athlete’s recorded choice.
Arciv product walkthrough · illustrated episode.