Interested in diving into research? Join ACM Research, an exclusive undergraduate program designed to introduce you to the world of research. You'll gain essential skills, access valuable resources, and receive mentorship from experienced researchers. Throughout the program, you'll develop sophisticated research ideas and work towards potential publication. Your journey will start with crafting compelling research questions and culminate in presenting your findings to industry and academic professionals at our symposium!
meet the directors
Oscar Lay
Research Director
Sahas Sharma
Research Director
experience
ACM Research helped me gain research experience and land a paid research position as an undergraduate.
I got to learn a ton about AI/ML/NLP that your courses won't teach you.
ACM Research has been the essential building block to help me become a researcher that I want to be.
The community at Research was welcoming and we continue to do the same for every new member. The relationships formed with all the participants and leads definitely stay forever!
Helix
Helix set out to predict the biological impact of missense mutations on protein growth by integrating structural bioinformatics with interpretable machine learning. Using the ThermoMutDB dataset of over 13,000 mutations we engineered a rich feature space to capture physiochemical, evolutionary, and 3D structural characteristics. With this dataset we developed a custom voting ensemble that utilized random forests, gradient boosting and SVM classifiers, which outperformed Siamese Neural Networks by 23% in classifying stabilizing vs destabilizing mutations. Our approach generalized well over a diverse set of protein families and also improved the interpretability of a reliable and faster mutation based discovery model.
Fingertip Fluency
The research project achieved its goal of advancing American Sign Language (ASL) fingerspelling and exploring its intricate nuances. With the comprehensive dataset provided by Google, we successfully developed cutting-edge machine learning models, utilizing state-of-the-art, novel architectures. Our innovative approach not only accurately interpreted the dynamic hand movements in ASL and deciphered their meanings but also surpassed the top-performing paper by an impressive 7%. By leveraging advanced models like SignExplainer and PoseNet, we set a new benchmark in the field.