Kevin Zhu

About

I am a Chemical Engineering MASc student at the University of Toronto and a graduate researcher in Prof. Benjamin Sanchez-Lengeling’s Chemical Cognition Lab at the Vector Institute.

My research is focused on applying methods in AI/ML towards problems in molecular and materials discovery! Right now I’m working on creating tools to augment the capabilities of experimental researchers through deep probabilistic models for molecules, guiding experiments and active learning workflows under uncertainty. I care about interpretability through incorporation of calibrated uncertainty quantification and post-hoc attribution methods, in order to justify and accelerate responsible AI for autonomous scientific discovery.

I spent two years as a synthetic and analytical chemist before pivoting to machine learning, applying scientific intuition towards my modelling and my teaching (despite the bitter lesson).

Outside the lab I raced for the Canadian National Dragon Boat Team through a grueling selection process, attaining 3× gold and 3× silver at the 2025 IDBF World Championships in Germany. I also raced with the Iron Dragons Paddling Club (1× gold, 2× silver, 3× bronze at the 2024 Club Crew World Championships in Italy). I was also President of the U of T Run Club from 2024 to 2025, leading the campus’ largest student organization: I made it official U of T Sport & Recreation programming and organized an ~80-runner charity race that raised $500+ for a pedestrian footbridge in Bolivia.

The Canadian National Dragon Boat Team crew mid-race in red Canada kit, paddles catching the water in unison. — Racing with the Canadian National Dragon Boat Team in Brandenburg, Germany.

Education

MASc, Chemical Engineering (Machine Learning) University of Toronto 2025 – Summer 2027 (expected)
Thesis: deep probabilistic & generative models for uncertainty-aware experimental research guidance. NSERC CREATE Trainee, Acceleration Consortium (AI & automation for materials discovery). Coursework: AI for Applied Chemistry & Chemical Engineering (A+); Chemical Data Science & Engineering (A+); Fundamentals of Chemical Engineering (A−).
HBSc, Pharmaceutical Chemistry University of Toronto · Focus in Practical Chemical Analysis 2021 – 2025
Dean’s List, all years. Two research theses, both graded A.

Research

Topic

Graduate Researcher, Chemical Cognition Lab Prof. Benjamin Sanchez-Lengeling · University of Toronto & Vector Institute 2025 – present
- LatentPockets. Generative models that learn protein pocket representations from protein–ligand data, helping experimental researchers with drug target selection in filling structural biology data gaps. Oral Talk, CSC 2026, Toronto.
- Deep kernel learning for polymers. Benchmarked graph neural networks with spectral-normalized Gaussian processes for low-data polymer property prediction across 6 datasets and 13 targets, fusing molecular graphs with process conditions. Attributed structure vs. process contributions via integrated gradients and GP-SHAP.
In collaboration with the Structural Genomics Consortium (SGC) and NC State University.
- VAEs
- Diffusion models
- GNNs
- Gaussian processes
Researcher, Medicinal Chemistry Self-Driving Lab Acceleration Consortium May – Aug 2025
- Evaluated chemical representations and clustering of Enamine building blocks for use in Reaction-GFlowNets.
- Proposed and validated automatable synthetic experiments.
- Clustering
- Cheminformatics
- GFlowNets
- Self-driving labs
Research Assistant, Taylor Group Prof. Mark S. Taylor · University of Toronto 2024 – 2025
Independently developed a photoredox route to threose nucleosides: 47% yield over 5 steps, more than 20% above prior methods. Thesis grade: A.
- Photoredox dehomologation of ribonucleosides (5′-carbon excision) to bypass lengthy threose nucleoside triphosphate syntheses.
- Explored three excision routes: reductive dehydroxymethylation (NHP ether), decarboxylation (NHP ester), and decarbonylation (aldehyde).
- Photoredox catalysis
- Nucleosides
- Total synthesis
Research Student, Peng Group Prof. Hui Peng · University of Toronto 2023 – 2024
Synthesized combinatorial organophosphate ester libraries for high-throughput structure–activity and toxicity screening, characterized by quantitative ³¹P-NMR and MS. Thesis grade: A.
- Libraries served as training data for environmental machine learning in CoSyER (see publications).
- Ran L-FABP1 binding assays by fluorescence displacement spectroscopy.
- Combinatorial synthesis
- qNMR
- Toxicology
Research Assistant, Batey Group Prof. Robert A. Batey · University of Toronto May – Aug 2023
- Identified acylating electrophiles selective for primary over secondary amines via competition experiments, for multicomponent couplings.
- Measured selectivity ratios by quantitative ¹⁹F-NMR using ortho-fluorobenzoyl electrophiles.
- Chemoselectivity
- qNMR

Selected projects

LLM-Guided Autonomous Optimization Agent Open Scientific Intelligence Hackathon Oct 2026
- An autonomous agent that uses LLM reasoning to steer optimization campaigns for materials discovery. Manuscript in preparation.
- LLM agents
- Optimization
Turbidity Forecasting & Anomaly Detection Course project · University of Toronto Jan – Apr 2026
- TCN forecasting and anomaly-detection benchmark on 10+ years of raw Ontario water treatment plant data, aimed at reducing excess upstream coagulant dosing.
- Traced its predictive ceiling to low target variance (0.0015 NTU²) against a ±0.02 NTU sensor noise floor.
- Time series
- Anomaly detection

Publications & Conferences

Type

Jun 2026
Journal
CoSyER: A High-Throughput Data Generation Approach for Environmental Machine Learning

Miller, J. M.; Cheng, Z.; Nair, P.; Wang, X.; Zhu, K.; Peng, H. · ChemRxiv preprint; under revision at Environmental Science & Technology.

DOI 10.26434/chemrxiv.15005434/v1
May 2026
Oral
LatentPockets: Addressing Structural Biology Data Gaps by Learning over Protein Subpockets

Canadian Societies for Chemistry and Chemical Engineering Conferences and Exhibition · Toronto, ON.
Mar 2025
Oral
Photoredox-Enabled Dehomologation of Ribonucleosides Towards Threose Nucleosides

Southern Ontario Undergraduate Student Chemistry Conference · University of Guelph.
Apr 2024
Oral
Combinatorial Organophosphate Synthesis for High-Throughput Structure-Activity Relationship

Southern Ontario Undergraduate Student Chemistry Conference · Western University.
Aug 2024
Poster
Application of Photocatalytic Dehydroxymethylation of Nucleosides Towards Nucleoside Triphosphate Synthesis

Departmental research symposium.
Mar 2024
Poster
Combinatorial Organophosphate Synthesis for High-Throughput Structure-Activity Relationship

Departmental research symposium.
Aug 2023
Poster
Chemoselective Acylations of Primary over Secondary Amines via Competition Experiments

Departmental research symposium.

Awards & Scholarships

Jan 2026 McLean Foundation Graduate Scholarship in Science and Technology$6,000
Apr 2025 Acceleration Consortium Research Award in Accelerated Discovery$8,500
Apr 2024 DuPont Canada Inc. Research Fellowship Award$7,500
Apr 2024 New Member of the Year Award
U of T Clubs & Leadership Development; selected from 10 nominees.
Mar 2023 DuPont Canada Inc. Research Fellowship Award$4,500

Teaching

Machine Learning Teaching Assistant CrossTALK AI-for-drug-discovery program · Data Science Institute, University of Toronto Jun 2026
- Debugged student ML code in Python and clarified concepts one-on-one and in small groups.
- Advised on building and tuning drug-candidate prediction models.
Program: CrossTALK: Cross-Training in AI and Laboratory Knowledge for Drug Discovery
Teaching Assistant — CHMB41: Organic Chemistry I University of Toronto Jan – Apr 2026
- Ran and graded group problem-solving tutorials and online quizzes on organic chemistry fundamentals and mechanisms.
Teaching Assistant — APS110: Engineering Chemistry & Materials Science University of Toronto Sept – Dec 2025
- Prepared 4 hours of weekly review and problem-solving tutorials for 80 first-year engineering students.
- Led active-learning sessions, graded assignments and exams, and invigilated.