Projects
Open-source work spanning protein language models, transformers, and multi-modal deep learning. More on GitHub.
FM4Life
Structured knowledge bundles ("skills") that let AI coding agents work with life-science foundation models like ESM, AlphaFold, RFdiffusion, DiffDock, and scGPT out of the box.
- AI Agents
- Foundation Models
- Life Sciences
finetune-esm
Scalable fine-tuning of ESM-2 protein language models with distributed training (FSDP/DeepSpeed via Ray) and parameter-efficient techniques such as LoRA.
- PyTorch
- Protein LMs
- LoRA
- Distributed Training
image_tabular
Python library for training deep learning models that combine image and tabular data in a single network, built on fastai. Used for the SIIM-ISIC melanoma classification challenge.
- Deep Learning
- fastai
- Computer Vision
protein-transformer
A step-by-step, from-scratch transformer implementation for antibody classification, with training, tuning, and evaluation included.
- Transformers
- PyTorch
- Antibodies
ReceptorAI
AI-driven T-cell receptor (TCR) matching and antigen discovery platform for immunotherapy research.
- Immunology
- TCR
- Machine Learning
citeseq_autoencoder
Autoencoder-based integration of single-cell CITE-seq multi-omics data. Companion code and notebooks for the blog tutorial and video series.
- Single-cell
- Autoencoders
- Multi-omics
DeepUFC2
Predicting UFC fight outcomes with deep learning on web-scraped fight and fighter statistics.
- Deep Learning
- Web Scraping