July 1, 2025
Day 26 - Model Accuracy & Presentation Kickoff
What I Learned
Today was a productive day in the lab as I continued making steady progress on the AI cardiovascular diagnosis project. My main focus was reviewing the performance of the models I had trained earlier—especially the 1D CNN—and analyzing its validation accuracy, which reached around 90%. I also helped ensure that the dataset pipeline was running smoothly by verifying the quality of the FFT-transformed input data and confirming that all classes were properly balanced. In addition to technical work, a significant part of the day was spent beginning preparations for our presentation. As a team, we started outlining the structure of our slide deck and deciding on key accomplishments to showcase. We discussed how best to present our objectives, model architecture, and results in a way that’s accessible to a broader audience. I also began drafting my individual talking points and gathering visuals, such as confusion matrices and architecture diagrams, to include in the slides. Overall, today marked an important shift from purely technical development to shaping how we’ll communicate our impact during the presentation.
Blockers
No Blockers
Reflection
Today marked a shift in focus during my internship—from building models to thinking about how to present them. Most of the morning was spent reviewing the 1D CNN model I had trained earlier. Seeing it hit around 90% validation accuracy was a rewarding moment. After spending so much time tuning the data pipeline and adjusting the architecture, it felt good to step back and see measurable results. But beyond the numbers, today was also the beginning of something new: presentation preparation. As a team, we started planning how we’d communicate our entire journey—what we built, why it matters, and what we learned along the way. We brainstormed the structure of our slides and began deciding which visuals would tell the story best. I found myself thinking not just like a developer, but like a storyteller. It reminded me that good research isn’t just about building strong models—it’s also about making others care about what you’ve built. Today helped me start making that shift.