Aryan Sharan Guda
Generative models for protein design
I am a Master’s student in Quantitative Biology and Bioinformatics at Carnegie Mellon University, working on my thesis in Prof. Lei Li’s lab. Before CMU I did a double major in Biotechnology & Bioinformatics and Electrical Engineering at IIT Hyderabad.
My work is on generative models for protein design, and on the pipelines that judge what those models produce. Most of what I do day to day is that evaluation side: whether a design folds as intended, whether it is physically plausible, and whether a set of designs is diverse rather than one shape repeated. Before this I worked on multimodal representation learning for drug and target data.
Affiliations
Research
Lei Li Lab, Carnegie Mellon University
Jun 2026 – PresentEnantioselective ligand-binding protein design, in collaboration with the Yang Lab at UC Santa Barbara. I work on the evaluation side, building LigandMPNN and ESMFold pipelines that score generated designs on designability, diversity, steric clashes, Ramachandran quality, and other structural measures. I also train models and test new design strategies. One architectural modification I proposed and tested has since been adopted into the project’s pipeline.
I am also developing a framework for dual-target binder design that does not depend on any one model: binary and ternary complex evaluation, structural compatibility metrics, and inference-time steering, applied across current de novo design models.
Chiu Lab, UPMC Hillman Cancer Center
Jan 2026 – May 2026Multimodal representation learning for drug-target interaction. I reimplemented CLOOME, a contrastive model that pulls Cell Painting morphology, drug SMILES, and protein sequence into one embedding space, and reproduced its core results. I then built a benchmarking pipeline to compare it against baselines. The work started with a review of deep learning methods for Cell Painting, including ChaMAE-ViT, masked autoencoders, and Cell Painting CNNs, to decide what was worth building on. End-to-end PyTorch, scaled on ACCESS Bridges-2.
Teaching
Applied Cellular and Molecular Biology, Carnegie Mellon University
Aug 2026 – PresentGraduate teaching assistant. I design and grade assignments and exams, and hold office hours.
Publications
-
Towards Federated Learning Across Biobanks: Prototype Software from the 2026 Carnegie Mellon University–NVIDIA Hackathon
BioHackrXiv preprint, March 2026
A report on ten federated-learning prototypes built at the CMU and NVIDIA Federated Learning Hackathon, all on NVIDIA FLARE. The prototypes cover genome-wide association analysis, histopathology harmonisation, pangenome construction, ancestry deconvolution, rare disease stratification, cancer subtyping, polygenic risk aggregation, and multimodal fusion.
News
-
Oct 2026
The slides for our AlphaFold 3 paper presentation in Generative AI for Biomedicine at CMU are online as live browser slides. The architecture is drawn as click-by-click animations, the results use the paper’s own figures, and an appendix holds the backup slides. Made with Julia Liu on Ramith Hettiarachchi’s slides template. Source.
-
Oct 2026
An AlphaFold 3 architecture explainer is live: a drillable map from tokenization through the recycling trunk and diffusion sampler to the confidence head, with a teaching trace that walks a small peptide, ligand and ion complex across 4 recycles, 48 Pairformer blocks and 200 sampler steps. The trace runs on deterministic toy values, not AF3 weights or predictions. Built for a paper presentation in Generative AI for Biomedicine at CMU, on top of Ramith Hettiarachchi’s Architecture Explainer. Source.
-
Aug 2026
pLaTeX is live: a typesetting game that runs a real TeX Live 2025 engine in the browser, compiled to WebAssembly. It grades you by compiling both your source and the target and comparing the two pages pixel for pixel, so anything that typesets identically counts. Four modes, from an untimed curriculum to a three-minute run. Source.
-
Aug 2026
Started as a graduate teaching assistant for Applied Cellular and Molecular Biology at CMU.
-
Jun 2026
Joined Prof. Lei Li’s lab at CMU, mentored by Ramith Hettiarachchi, and continuing there through the fall for my thesis.
-
Mar 2026
Our report from the CMU and NVIDIA Federated Learning Hackathon is up on BioHackrXiv, covering ten federated-learning prototypes for biomedical data.