GEMASTIK Project

Overview

GEMASTIK is Indonesia's national ICT competition. Our entry combined an applied machine-learning model with a focused product narrative, judged on both technical rigor and real-world framing.

Approach

We treated the model as a product from day one: a reproducible training pipeline, an honest evaluation harness, and a thin serving layer that made the results tangible to non-technical judges.

Highlights

• A clean data-to-evaluation pipeline that anyone on the team could rerun
• A serving API that turned model output into a demoable experience
• A written rationale connecting metrics back to the problem statement

Reflection

The project sharpened how I communicate ML work — leading with the decision a model enables, not the architecture behind it.