Programming Work

Problem: 

Our scientific instrument to measure in-plane thermal conductivity must output sinusoidal heating to a sample, measure temperature from 6 thermocouples, display live data, and have the ability to control test paramters mid-test, all while remaining an accessible program to a researcher with no programming experience.

Action:

We decided early that software would be split into subsystems, being the Thermocouples, the Heaters, and the Web Frontend. Connecting each subsystem would be a database for full traceability in recording all data, with additional HTTP requests used where required.

Python was chosen for the backend connections for its ease of use and pre-existing examples for the Pico-Talker and TC-08 Handler. SQLite was chosen as the central database for its simplicity as a single-machine database. And a Next.JS frontend was built as a team member had previous experience with the framework, and a web-based frontend has the advantage of being easy to learn and use for researchers with no experience in a terminal window.

Results:

The program was successfully able to start, monitor, end, and analyze tests, with an easy to understand UX guiding the laboratory technician. Our additional work to surpass the stretch goals set by our client in creating this interface helped our team win 1st Place in our track over 7 other teams, voted unanimously by experts and alumni.

The project is viewable on Github. 

Lessons Learned:

• Dividing software work into subsystems and collaborating to ensure compatibility

• Designing a large-scale software solution from scratch

• Full-stack web app development

Problem: 

Micro-needle Patches manufactured at Vaxess can only be inspected by trained engineers looking at print images, or in final stages where resources may have been invested into the product that should have failed multiple steps prior. Existing Pre and Post-Print Machine Vision Inspection on the Production Line is primitive and requires a total overhaul.

Examples of Images taken of Molds before and after different print rounds

Action:

I fine-tuned the proprietary Machine-Vision software to be far better at recognizing the "blobs", or individual print fills in each mold indent. 

Each blob could then be analyzed for different statistics, where it's size, eccentricity, location, "bubbly-ness", general cleanliness, etc. could be used to determine if the print was a success or not, and even be given a grade of quality. The most progress in this area was done in predicting "Tip Length", a factor that was measured a day after a print and often disqualified batches that had already had time invested onto through human inspection and packaging. That time waste is now avoided through this machine vision inspection, that can predict tip length based on blob images with an R^2 value of .878.

Examples of Overlays added to Mold Images to assist in in-production inspection

Additionally, failures in printing, such as extremely large blobs are automatically identified, and a print round is automatically stopped for operator inspection. And all statistics were able to be exported to .csv files for easy data analysis.

Examples of Overlays added to Mold Images to assist in in-production inspection

Results:

The machine vision software was still in development and fine-tuning as I left the co-op, but the version shown was installed and helped greatly in lowering the burden on the technician to manually inspect each image with no guide. Further development should allow for statistics generated to give an easy to understand grade of the print, as well as assisting operators in diagnosing which specific print settings should be adjusted.

Lessons Learned:

• Working on Proprietary Software alone with limited support, figuring out quirks and solutions on my own

• Designing software interfaces to present only necessary information in a clear, direct manner

• Fine-Tuning Machine Vision Algorithms

Game development is a hobby of mine that I haven't had enough time for. I'm very passionate about the craft, but have had to shift my priorities elsewhere. Above in my itch.io page there are the few finished prototypes/tests/game-jam entries that I've completed. Given more time I'd love to dig deep on a larger projects I've been ideating on.