This project integrates microelectronics with machine learning to develop a portable chemical gas-sensing platform invariant to real-world environmental noise. Traditional gas sensors are limited by humidity, electrical noise, and device-to-device variation, which reduce measurement reliability and prevent consistent data acquisition for machine learning. This research addresses these issues by creating a sensor chip using carbon nanotube field-effect transistors (CNFETs) and investigating both hardware and software approaches to improve device robustness. Through systematic variations in environmental conditions and gas concentrations, this project constructs the first microelectronics-based, large-scale gas sensing dataset operating at room temperature. Machine learning techniques, such as convolutional neural networks and pattern recognition models, will be utilized to distinguish gas-specific responses from environmental interference. I will also implement physical modifications, such as hydrophobic coatings and material changes, to reduce sensitivity to external disturbances. The project investigates the limits of computational mitigation by combining experimental device design with machine learning analysis, identifying which non-ideal behaviors can be learned and which require hardware-level intervention. There is currently no portable gas sensing chip for general use, and creating one has many potential applications, ranging from environmental monitoring and industrial safety to healthcare diagnostics and smart agriculture. Precise detection of gas leaks or emissions is essential for optimizing energy efficiency and reducing waste. This research establishes novel methodologies for collecting and evaluating extensive datasets from physical sensors, providing significant insights into how hardware design and machine learning can be united to enable accurate and resilient chemical sensing platforms.

I am excited to participate in this SuperUROP and explore the intersection of microelectronics, machine learning, and energy-efficient computing. Through my previous research, I have become very interested in a fundamental challenge facing modern technologies: how to scale AI within the energy, power, and hardware constraints of real-world systems. This project provides an opportunity to investigate this topic while gaining deeper experience from both hardware and software perspectives. I look forward to strengthening my communication skills through SuperUROP by learning from experts, engaging with other students, and contributing to innovative and impactful research.