Albert Lee
I build statistical models, the data systems under them, and AI tools that scientists and engineers use every day.
Senior Staff Bioinformatics Scientist at Guardant Health, which detects cancer from a blood draw. PhD in Biomedical Informatics from Columbia. Since 2011 I have worked on biological data analysis, statistical modeling and research infrastructure, and lately on AI tools for the people doing that work.
Selected work
- PhD on Bayesian multilevel models of immune response to Marburg virus (Columbia).
- Shipped a Bayesian prediction model at Myriad Genetics that scores 110,000+ clinical samples a year.
- Wrote the statistical analyses behind FDA clinical-validation reports for blood-based cancer tests.
- Built Guardant's research data warehouse on Snowflake and dbt; 100+ scientists and analysts query it.
- Airflow, Nextflow, Python and SQL pipelines; a postmortem for every incident, and a check so the same failure raises an alert next time.
- Wrote The DiCE Principle (2024): results are only reliable when you manage their Data, Code and Environment.
- Built an LLM assistant in Slack used by 200+ colleagues.
- Started my company's AI coding program as a self-built pilot, and helped grow it into an enterprise rollout across R&D and engineering.
- At home I build my own agent system: critic agents that challenge my decisions, calling 200+ small tools I wrote, each in its own repo.
Things I made that you can try
- Learning games I built for my son when he fell for numbers.
- A personality test with no paywall, because the "free" one asked me to pay for my own result.
- Essays and public notes.
Selected research
- Co-author of 10 peer-reviewed papers, including Nature (2014, 2017), Cell (2018), Cancer Cell, PNAS and Immunity.
- First author: the first transcriptome of the Egyptian rousette bat, the natural reservoir of Marburg virus (BMC Genomics, 2015).
- Genomic surveillance of the 2014 Ebola outbreak (PLoS Currents, 2015).
Get in touch
Message me on LinkedIn about Bayesian modeling, research data infrastructure or AI developer tools. Tell me the problem you are working on and what you would like to discuss.
Off the clock
I study close-up magic the way Juan Tamariz teaches it: the secret matters less than the moment it leaves with the person across the table. I'm bilingual in English and Korean, and learning Japanese one kanji at a time.