Cornell Computer Systems Lab

May 2025 - Present

Working under Professor José F. Martínez.

Developing parallel computing and high performance computing labs for ECE 6750: Advanced Computer Architecture, a graduate course in computer architecture. Using C, OpenMP, IntelMKL and, the pthread POSIX library. Current labs involve hardware-aware parallelization of matrix operations and sparse matrix operations.

Course Assistant

January 2025 - Present

Current: Course Assistant for CS 2800: Discrete Structures (August 2025 - present)

Past: Course Assistant for CS 2110: Object Oriented Programming and Data Structures (Jan - May 2025)

Cornell Prism Laboratory

May 2025 - August 2025

Working under Professor Sainyam Galhotra.

Concept bottleneck machine learning models provide increased explainability and resiliency to spurious correlations as opposed to traditional black box learning models, however this can come at the cost of accuracy. I worked to improve the accuracy of Post-Hoc concept bottleneck models (https://arxiv.org/abs/2205.15480) and Label-free concept bottleneck models (https://github.com/Trustworthy-ML-Lab/Label-free-CBM) on the Living17 dataset, a subset of ImageNet using LLM-based concept addition. Achieved a 40% worst-group accuracy increase through addition over baseline concept set. Used CLIP, Python, HuggingFace LLMs, and Pytorch.

Cornell SonicMEMS Lab

May 2024 - June 2024

Selected components for PCBs and designed PCBs for a solar microrobot. Completed schematics, and layout using Fusion 360. Boards assembled included a solar cell breakout board and MM101 evaluation board. Boards designed included an amperage breakout board and a bluetooth low energy module. Assembled PCBs by hand and with a Neoden YY1 pick and place machine. Built a 3D-printed vacuum stencil fix machine to aid in applying solder paste to PCBs.

Boyce Thompson Institute at Cornell University

Jun 2023 - August 2023

Completed during the BTI 2023 internship. This project used RNA sequencing (RNA-Seq) data from different fruit tissues at different developmental stages of the F1 cross between the M82 domestic tomato and the *S. Pennellii* wild tomato.

RNA-Seq reads were processed and mapped to the genomes of the two parents. Based on the mapping quality to each parent genome, the reads were assigned to one of the parents, and raw and normalized (FPKM) read counts for each of the two alleles were calculated for all genes. Gene ontology analysis and identification of ASE genes was then performed, leading to the identification of 24 significant ASE genes at all stages.The identification of these candidate genes could allow for future functional analysis that will deepen our understanding on how cross breeding with wild species could improve flavor and nutritional quality in cultivated tomatoes.