Advisor: Connor Coley (MIT ChemE)- Designing novel algorithms to enable the application of continuous optimization techniques to discrete, non-enumerated spaces (e.g., virtual chemical libraries)- Researching model-free optimization techniques for efficient traversal and search through virtual libraries based on a given property objective (e.g., docking score, ADME properties, etc.)

Advisors: Connor Coley (MIT ChemE), Eugene Shakhnovich (Harvard CCB)- Developed software (MolPAL) to accelerate ultra-large (> 100M) molecular docking screens using Bayesian optimization, machine learning, and cheminformatics- Established Design Space Pruning (DSP), an algorithmic extension to active learning that can reduce overhead costs associated with surrogate ML model inference by 60-80% while retaining optimization performance- Lead developer of Chemprop, a Python package for machine learning using directed-message passing neural networks. Rewrote the entire codebase from the ground up as Chemprop2, increasing the speed by 40% while reducing the footprint by 50%- Refined roughness index to enable quantitative comparison of QSPR surface roughness across multiple representations (ROGI-XD). Utilized ROGI-XD to better understand the performance of unsupervised chemical representations from LLMs, VAEs, and pretrained GNNs, in various QSPR benchmark tasks.- Collaborated with experimentalists to utilize MolPAL in the identification of novel antimicrobial compounds

Joint graduate student in both the Betley and Jacobsen groups in the Department of Chemistry and Chemical Biology at Harvard University. Aided in the development of a reaction for enantioselective C-H activation via nickel catalysis and studied the ability of hydrogen bond donor catalysts to affect the reactivity of both terminal metal oxos and terminal metal nitrides

- Integrated active learning software (MolPAL) into drug discovery workflows to accelerate the screening of large compound sets (10-100M) using computational docking and machine learning- investigated the extension of docking-based active learning workflows to molecular dynamics-based RBFE virtual screens- designed ML models to predict compound relative binding free energies directly from output docking pose through multimodal training using the Boltzmann-averaged trajectory and calculated RBFE

At Private Prep, I work directly with students to conduct remote tutoring sessions for introductory chemistry, AP chemistry, and AP Computer Science. My work ranges from directly teaching the subject matter to working on chemistry problem-solving skills to more general concepts of test-taking and test-preparation

Undergraduate Research Assistant in the Knowles Research Group where I worked to develop a novel, photocatalytic method for olefin hydroamination that was published in a high-tier, academic chemistry journal. Further aided in the elaboration of this reaction into an enantioselective variant that was also published in a high-tier chemistry journal.

Private tutor for students of Princeton Tutoring. Conduct private tutoring sessions with students in varying subjects. Record progress and maintain client satisfaction through written recaps. Schedule sessions by reconciling both client and personal obligations

The BCMP Summer Scholars Program is a summer research program through the Department of Biological Chemistry and Molecular Pharmacology at Harvard Medical School. Students conduct independent research projects under the individual guidance of a faculty mentor and attend weekly seminars where various faculty in the BCMP present on their group's research. During this time, students function as full members of their respective labs, attending lab meetings and other functions, and ultimately conclude with a written manuscript of their research project. This research then culminates with an oral presentation.