This October, I had the incredible opportunity to present my research poster, “Enhancing Failure Detection in Semiconductor Manufacturing using Balanced Random Forest Model,” at the MIT Undergraduate Research Technology Conference (URTC) 2025 — one of the world’s premier venues for undergraduate research in science, technology, and engineering.

The reason why I chose this topic was because I was learning basic ML and I thought it would be a great idea to write a paper , that would help me research more about various models and learn . I have some interest in semiconductors and circuits, hence I chose this area for my research. The project I presented began not as a competition entry, but as a curiosity. I was fascinated by how something as invisible as data imbalance could distort our understanding of real-world systems. Semiconductor fabrication felt like the perfect test bed , a field where precision is everything and even a 1% error can ripple through billions of devices.

Modern semiconductor fabrication is a marvel of precision — a single microscopic defect can compromise entire batches of chips. Even a 6 percent wafer failure rate can cause millions of dollars in yield loss. Yet, because failure data are so rare, traditional machine-learning models tend to overlook them.
My research set out to solve this imbalance problem: Can we design AI models that “care” equally about rare failures and common passes?

Designing the poster turned out to be its own experiment in communication. I had to distill hundreds of lines of Python, 591 wafer features, and months of iteration into a few visuals and simple phrases that anyone ,engineer or not could connect with.
It forced me to think how I tell stories with data: not by overwhelming the viewer with numbers, but by walking them through the why behind the results. In doing that, I realised that research isn’t just analysis ,it’s empathy. You have to imagine how someone else might interpret your work, where they might get confused, and what might spark their curiosity.

MIT URTC taught me more than data science could alone.
I learned that:

• Clarity beats complexity — people connect with your reasoning, not your equations.
• Interdisciplinary thinking is the future — AI isn’t separate from manufacturing, sustainability, or ethics. It’s the bridge between them.

Presenting there gave me confidence that research isn’t defined by age or institution but by defined by curiosity and persistence.This project marks the start of a broader journey toward AI-driven semiconductor reliability. Future work will test cost-sensitive deep-learning models and integrate adaptive techniques to handle concept drift in modern fabs. For me, MIT URTC 2025 was more than a conference ,it was a glimpse into a future where data science and hardware engineering converge to build cleaner, smarter, and more efficient technologies.