Research

My research focuses on building ML systems for scientific problems — particularly at the intersection of AI, chemistry, and structural biology. I’ve worked across protein design, RNA folding, molecular modeling, and LLM-driven bioengineering.

University of Toronto — Matter Lab (Aspuru-Guzik Group)

Undergraduate Research Project — Unhobble Science: Building a Model-Adaptive Layer in Chemistry Agents | September 2026 – Present | Toronto, ON Advised by Professor Alán Aspuru-Guzik and Professor Varinia Bernales

• Studying how to unhobble a model and move knowledge between harnesses: what is transferable across scaffolds, what is scaffold-specific, and what turns out to be unnecessary altogether.

Acceleration Consortium — Alán Aspuru-Guzik Group

Machine Learning Research Intern — Agentic AI in Self-Driving Labs | May 2026 – Present | Toronto, ON

• Architected and implemented a full-stack platform connecting LLM agents with laboratory hardware and scientific software through validated, permission-scoped interfaces for self-driving laboratories.
• Deployed and served open-weight LLMs on local compute clusters, building inference infrastructure and tooling for researchers to run models within on-premise scientific workflows.
• Engineered closed-loop agentic workflows for experiment execution, data analysis, and adaptive decision-making, integrating physical hardware with automated tool execution and failure recovery.
• Continuing part-time after the summer internship on reliable agentic reasoning and decision making — making an agent’s reasoning trustworthy enough to act on.

University of Toronto — Chemspace Lab

Machine Learning Research Student | May 2024 – April 2026 | Toronto, ON

• Developed a data-efficient learning architecture by deriving symmetric representations from second-derivative Hessian and DFT matrices, accelerating the modeling of alchemical exchange invariant molecules by 2×.
• Investigated methods for designing chemical features for molecular energy curves by comparing interpolation-based 3D sorting with ML-based approaches, analyzing how ordering/representation of atoms affects downstream model interpretability.
• Implemented algorithms such as A*, locality-sensitive hashing, and genetic algorithms on molecular energy curve sorting. Developed a KNN classifier to group molecules for generating smooth 3D molecular energy curves.
• Explored how a generative Gaussian mixture model with cross-validation using PyTorch could predict low-energy conformers of 4-atom molecules.

Yale University — Gerstein Lab

Machine Learning Research Assistant — Protein Design | August – September 2025 | Remote

• Contributed to an inverse-folding based protein design paper (under review at Cell Systems), including designing evaluation experiments and writing sections of the manuscript and point-to-point rebuttal letter.
• Researched and implemented experiments to test model performance against protein design benchmarks (PepMerge for peptide design, SAbDab for antibody design) to address reviewer concerns.
• Analyzed sources of poor model performance on specific proteins in the CATH 4.2 test set, providing explanations to guide further refinement.
• Improved model performance by addressing benchmark limitations, including implementing new evaluation metrics informed by chemistry and biology background.

National University of Singapore — SERIUS Internship

Machine Learning Research Intern — Large Language Models (LLMs) | May – August 2025 | Singapore Supervised by Yeoh Jing Wui and Professor Poh Chueh Loo

• One of thirty students from North America selected for the 2025 SERIUS internship program at NUS.
• Defined the project scope and led end-to-end research on evaluating and fine-tuning LLMs for metabolite pathway engineering.
• Engineered a complete dataset pipeline by converting a newly published metabolite database into graph representations and generating chain-of-thought prompts to improve reasoning over bioengineering data.
• Designed and conducted experiments evaluating table-, graph-, and text-based fine-tuning using models such as Google Gemma and DeepSeek, benchmarking with BLEU, ROUGE, and custom biological reasoning metrics.
• Secured $8,000 CAD in research funding from the University of Toronto to support project execution.
• This work grew into Recovering Hidden Chemistry: Improving LLM Fine-Tuning on Metabolic Pathways, accepted to the NeurIPS 2026 Women in Machine Learning Workshop.

Pioneer Academics

Researcher — Computational Chemistry | May – September 2022 | Remote

• Modeled the secondary and tertiary structures of the HIV-1 RNA terminal loop hairpin, integrating computational chemistry and molecular structural analysis under the mentorship of Professor Lectka (Johns Hopkins University). Achieved a grade of A.
• Utilized Gaussian, GAMESS, and Spartan to optimize 3D RNA conformations and studied how hairpin folding influences key molecular properties, including polarity, energy, intramolecular forces, and interactions with other molecules.

University of Alberta — Brown Lab

Research Assistant — Computational Chemistry | October 2021 – July 2023 | Edmonton, AB

• Assisted in the research of the effects of distance and location of a charge on a HBDI chromophore molecule’s excitation energy in different solvent environments.
• Wrote scripts to automate DFT/Hartree-Fock calculations of molecules on Compute Canada with Bash.