01
Protein AI Active exploration
ProtDiffusion
A PyTorch implementation exploring diffusion-based protein structure prediction.
Zaid Yusuf
ML Research
Engineer
Scientific ML · Protein AI · AI Engineering
I work across the ML stack: mathematical foundations, model architecture, implementation, systems, and production—especially where scientific modeling, biological data, and reliable engineering meet.
01 / Selected work
Research meets implementation.
↗
02
Biological foundation models Open source
scGPT-mini
An implementation study of scGPT for single-cell transcriptomics.
03
Agentic AI Open source
BioAgent
A biology-focused tool-calling agent fine-tuned with SFT/LoRA on open-source language models.
04
AI Engineering Ongoing technical work
ML systems & inference
Implementation-led study of the systems choices behind useful, reliable LLM applications.
02 / Technical writing
Notes from the work.
Writing is where I make the machinery explicit: geometry, models, systems, and the decisions hidden inside an implementation.
Sep 2026
A working intuition for diffusion on SO(3)
Why rotations make a useful case study for respecting geometry in generative models.
Aug 2026
MHA, MQA, and GQA: the cache is part of the architecture
A compact comparison of attention variants through the lens of inference-time memory.
03 / About
I care about understanding a model well enough to change it with intent.
I’m drawn to problems where assumptions matter: geometric representations, biological structure, inference trade-offs, and the engineering required to make ML systems useful outside a notebook.
More about my approach04 / Contact