Therapeutic peptide representations
Developing SMILES-based and structural-ensemble representations for therapeutic peptides, with an emphasis on molecular properties, membrane diffusion, and design.
Computational biochemistry · Austin, Texas
Teaching machines to reason about molecules.
I am Aaron Feller, Charles W. Smith Jr. Graduate Fellow at The University of Texas at Austin. I build deep learning systems for biological and therapeutic discovery.
Peptide design
Protein engineering
Molecular discovery
PhD Candidate
UT Austin
01 / About
Biology is full of noise. My work finds the signal hiding inside it.
I work at the intersection of deep learning, knowledge representation, and molecular design. The throughline is simple: build useful models that remain connected to the physical systems they describe.
My research spans protein engineering, therapeutic peptide design, and small-molecule and natural-product discovery. I build molecular representations that make difficult biological questions more tractable.
02 / Research
From molecular language
to useful design.
Protein engineering
Evaluating protein encoders across deep mutational scanning data to understand where zero-shot prediction helps and fails. New high-performing models are in development.
Molecular discovery
Using chemical similarity and genome-scale bioinformatics to discover useful small molecules, natural products, and antibiotic peptide scaffolds.
03 / Publications
Research in print
A selection of peer-reviewed work spanning molecular machine learning, peptide chemistry, and natural product discovery.
Google Scholar- 2026Graph Learning on Ensembles of Cyclic Peptides: An Investigation of Molecular Ensemble ModelingICML Workshop on Graph Foundation Models
- 2026Scaling SMILES-Based Chemical Language Models for Therapeutic Peptide EngineeringJournal of Chemical Information and Modeling
- 2026Overestimating zero-shot fitness prediction: Broad benchmarks mask local failures and practical limitationsbioRxiv
- 2025Peptide-Aware Chemical Language Model Successfully Predicts Membrane Diffusion of Cyclic PeptidesJournal of Chemical Information and Modeling
- 2025Antibacterial Microcins are Ubiquitous and Functionally Diverse Across Bacterial CommunitiesNature Communications
- 2023Using alternative SMILES representations to identify novel functional analoguesPatterns
04 / In the lab & beyond
Building a generous technical community.
2022 — Present
Graduate Research Assistant
The University of Texas at AustinAdvancing AI-guided protein engineering with the Wilke Lab and collaborators across computational biology and molecular engineering.
2024 - Present
Research Collaboration
Novo Nordisk A/SContributed to research on peptide-specific molecular representations and predictive modeling for therapeutic peptide engineering.
2023 — 2026
Co-Founder & Organizer
Biology & Machine Learning SocietyBuilt a student-led forum for AI in biology and organized an international protein engineering hackathon with more than 40 teams.
2025 — Present
Community Contributions
Volunteered TimeContributing to and reviewing for Terminal Bench Science, supporting data analysis for Nucleate HQ, and mentoring post-bacc students.
05 / Toolkit
Methods & tools
01
Modeling
Transformers · GNNs · Transfer Learning · Meta Learning · Probabilistic Modeling · Model Evaluation
02
Scientific computing
Python · PyTorch · Lightning · RDKit · NumPy · Pandas · R
03
Infrastructure
HPC · Slurm · TACC · AWS · git · reproducible data pipelines
04
Communication
Technical writing · LaTeX · teaching · scientific visualization · conversational Spanish