Notebook

Short explainers, videos, and working notes. Written for curious scientists, whichever side of the bench they sit on.

  1. NotePinned

    Why this notebook exists

    A place for short explainers, working notes, and the occasional video on statistics, machine learning, and immunology.

  2. Video, 3:35

    DESeq2 for bulk TCR-seq: a field guide from read counts to off-target calls

    A 3.5-minute narrated video and a complete tutorial on using DESeq2 (PyDESeq2) to compare V-gene usage between CAR-T and control T cells: why raw percentages mislead, the five steps inside DESeq2, a…

  3. Video, 5:35

    What is a survival curve? How to read the most important chart in clinical research

    A 5-minute narrated video and an illustrated walkthrough of the Kaplan–Meier survival curve: time to an event, censoring, why counting fails, the step-by-step calculation, medians and confidence bands, the log-rank test, the…

  4. Video, 9:34

    How is GPT trained? From next-token guessing to RLHF, step by step

    A 10-minute narrated video and an illustrated walkthrough of the four stages behind GPT and ChatGPT: pretraining, supervised fine-tuning, a reward model trained on human rankings, and RLHF with PPO. Every loss…

  5. Video, 8:59Paper reading

    The Virtual Biotech: an AI drug company made of agents

    A 9-minute narrated video and an illustrated reading note on Zhang et al. (Science, 2026): how the multi-agent Virtual Biotech is built (a CSO orchestrator, eight specialist agents, MCP tools and a…

  6. Video, 5:39Paper reading

    AI in drug discovery: where do we really stand?

    A 6-minute narrated video and an illustrated guided reading of Bender et al. (Nature Reviews Drug Discovery, 2026): why AI’s clinical impact is still thin, where the value really lies, why data…

  7. Video, 6:31

    What is PCA? A visual guide, told with blood cells

    A 6-minute narrated video and an illustrated walkthrough of principal component analysis: variance and the spinning line, covariance, eigenvectors and SVD, loadings and scores, real PBMC single-cell data, choosing the number of…

  8. Video, 6:37

    What a p-value is, and isn’t, in 10 minutes

    A hand-drawn 6-minute video and an illustrated walkthrough: one clinical trial, a shuffle test, the t-test, α and power, why significant isn’t important, four classic traps, and what 20,000 genes do to…

  9. Video, 9:47

    What is a Transformer? A visual guide, told with one T-cell receptor

    A 10-minute narrated video and an illustrated walkthrough: tokens and positions, query, key and value, the attention formula step by step, multi-head attention, the Transformer block, ESM-2 guessing a hidden residue, and…