Experimental biologist by training, Rust tinkerer by choice.
Pipetting by day, compiler battles by night, implied volatility in between.
I work at the intersection of experimental biology and computation, with a focus on CRISPR screens, cancer biology, and immunology. I enjoy designing experiments, making sense of messy data, and building tools that make research a little less painful.
CRISPR screens · Cancer biology · Immunology · RNA biology & multi-omics
My research focuses on understanding the biology behind disease and identifying potential therapeutic vulnerabilities through a combination of experimental and computational approaches.
I build high-performance bioinformatics tools in Rust, improving established algorithms and developing new workflows with a focus on speed, scalability, and scientific accuracy.
| CRISPR & functional genomics | Immune repertoires |
|---|---|
| CrisprClamp — High-throughput FASTQ processing and guide counting for single- and dual-guide CRISPR screens. | ClonoForge — Rust implementation of pRESTO for TCR/BCR sequencing preprocessing and quality control. |
| MAGeCKrs — Optimized and extended MAGeCK implementation supporting guide counting, RRA, MLE, and copy-number correction. | ClonoTrace — BCR clonotype identification and clonal lineage analysis. |
| ZRust — High-performance DrugZ implementation for detecting drug sensitization and resistance. | ClonoDist — Optimized TCR distance calculations and repertoire similarity analysis. |
| RSBagel — Optimized BAGEL2 implementation for Bayesian gene-essentiality analysis. | ClonoShift — Detecting changes in TCR/BCR clonotype abundance across conditions. |
The goal isn't just faster code. It's faster code that still gets the biology right.
I use AI coding agents as part of my development workflow, experimenting with different models, independent cross-model reviews, and iterative testing.
Different models have different strengths and blind spots. I find that having them challenge one another can be useful, but the final judge is always the code, the benchmarks, and the underlying biology.
- Respect the assay: High-depth sequencing won't resurrect poorly lysed cells.
- Deterministic always: Same input, different output? Straight to jail.
- Trust, but verify: AI can write the code, but it doesn't get to grade its own homework.
- Ground truth: P-values can be flattered; raw reads and cash balances don't lie.
- Respect the tail risk: Don't bet the farm on unhedged spreads or unvalidated sgRNAs.
Always happy to talk science, software, or markets.