Adeoluwa Adekanmbi
MIT Department: Electrical Engineering and Computer Science
Faculty Mentor: Prof. Paul Liang
Undergraduate Institution: Howard University
Website:
Biography
Adeoluwa is a rising senior majoring in electrical engineering at Howard University. Growing up in Nigeria, his diverse upbringing instilled in him a profound understanding of both privilege and disparity, shaping his perspective on global challenges. Adeoluwa is driven by a commitment to apply Artificial Intelligence (AI) in tackling global challenges and enhancing everyday life. His interests span AI-powered solutions for climate resilience, renewable energy, and human-centered technologies that improve quality of life. At MIT, Adeoluwa works under Prof. Paul Liang, integrating smell perception intoAI by modeling chemical signals from gas sensors using machine learning methods, with use cases such as allergen detection in food. He envisions pursuing graduate studies at the intersection of AI and sustainable innovation, aiming to develop technologies that serve both people and the planet.In his free time, he enjoys playing the piano, connecting with others, engaging in sports, and exploring creative ideas.
Abstract
Advancing Machine Olfaction for Real-World Allergen Detection
Adeoluwa Samuel Adekanmbi1, Paul Pu Liang2
1Department of Electrical Engineering and Computer Science , Howard University
2Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology
Machine olfaction—teaching artificial intelligence systems to recognize smells—has critical applications in food allergen detection and environmental monitoring. A key challenge is that existing datasets and models focus on isolated, pure substances, while real-world smells often arise from complex mixtures. We address this gap by extending the SMELLNET framework, the first large-scale dataset and modeling benchmark for smell recognition using portable gas sensors. Specifically, we curate a new dataset of food mixtures containing peanut allergens and develop improved machine learning models for mixture-based smell recognition. We benchmark these models under both controlled and real-world conditions, demonstrating improved mixture classification accuracy and robustness to environmental variability. Preliminary real-time inference results show the model confidently detecting peanuts in a mixture (~80% probability) while maintaining below-threshold scores on a non-peanut sample (~42%), suggesting strong potential for future deployment. Our work advances machine olfaction toward scalable, real-world applications and contributes new resources, modeling strategies, and benchmarks to the field. 