Right now
Heads-down on two things: the next post in the Silicon to Systems series (part 2 — adders and arithmetic), and the first of eight flagship interactive explainers laid out in my brand strategy — "Attention, But You Can Scrub Through It."
Day job: still leading ML systems work — production LLM orchestration, evaluation harnesses, the usual.
Applying to
The Doctor of Health Informatics program at UTHealth McWilliams School of Biomedical Informatics. Application due March 1, 2027 for an August 2027 start. Open to senior ML / applied-research roles at Anthropic, Google, Apple, and a handful of health-AI-adjacent companies in parallel — see the About page if you want to talk.
Learning
- Frank Harrell's Regression Modeling Strategies, slowly. The chapter on splines is changing how I think about feature engineering.
- The recent wave of mechanistic-interpretability work coming out of Anthropic and EleutherAI — Olsson et al. on induction heads, the more recent SAE work.
- R, working level — enough to read clinical-research code and write the companion notebook for the eventual survival-analysis post.
Reading queue
- Causal Inference: The Mixtape (Cunningham) — re-reading the diff-in-diff and synthetic-control sections.
- The Book of Why (Pearl) — for the third time. New things stand out every pass.
- Anthropic's recent papers on interpretability and faithfulness of chain-of-thought.
Not doing
Saying no to: speaking gigs that don't align with the explainer brand, contract work that would pull me off the writing cadence, social-media drift. The writing is the asset; everything else is friction.
This page is updated roughly monthly. Last touched: May 26, 2026. This idea is borrowed from Derek Sivers's /now page movement.