October 21, 2025 • By Tyler Vo
Student ML • FastAI Machine Learning
This term, I’ve been tightening the loop between coursework and production. In the studio, I’m standardizing molds, finishes, and packaging so every shipment looks intentional. In the lab, I’m building compact models for practical problems—prioritizing clarity, versioned experiments, and results that non‑technical audiences can use.
The website you’re browsing is part of that discipline. Pages now share the same masthead, navigation, and external stylesheet for a consistent, professional feel—because presentation matters as much as the analysis behind it. I’m also documenting decisions to make the process reproducible and future‑proof as new projects are added.
What’s New
- R+E Made: nicer unboxing and steadier curing times.
- VoAI Fusion: ML for cost behavior & inventory—now with FastAI.
- Web Refresh: centered layout, rollover nav, and a shared external CSS.
FastAI
I’m using FastAI to get to strong baselines quickly. For vision work, transfer learning with ResNet backbones makes short iterations possible; for tabular tasks, FastAI’s API helps balance regularization and interpretability. The emphasis is on usefulness— models that support decisions like inventory planning, defect triage, and timing.
Upcoming demos will include a small finish‑defect classifier with a confusion matrix and precision/recall, plus a tabular model that links studio capacity with order timing. I’ll publish metrics on the portfolio page as these stabilize.