02_ / content

Work I’m proud of

I share it in case something here sparks an idea, or gets you past a problem you're stuck on.

Read

  1. From notebook to post on the DGX Spark

    How a notebook on the DGX Spark becomes a post on this site, and why live HR records never leave the machine.

    r&d note

  2. Building a synthetic workforce of 5,000 people that holds up against public benchmarks

    Every example on this site runs on synthetic data. This is how I generate it, and how I test that it behaves like a real organisation.

    article

  3. Holding text for a written post

    This page stands in for a post that is only writing: no data, no code and no figures, and nothing on it asks for them.

    article

  4. A local language model on the DGX Spark

    Installing a local model on the Spark, and drafting a first post with it from the notes on this site.

    r&d note

  5. Fine-tuning a small model on this site's own posts

    Teaching a local model the site's house style with LoRA, and reading what its training curve shows.

    r&d note

everything to read

Watch

  1. Fuzzy matching two lists of names

    Two HR extracts with no shared key. I score how alike the names are, pick a threshold, and check what lands just under it.

    video, 6:40

  2. Survival curves for tenure

    Kaplan-Meier in a few lines of Python: how long people stay, and where the curve drops.

    video, 4:12

  3. An org chart from a flat file

    A manager ID column is enough. I rebuild the hierarchy from a flat extract and draw it.

    video, 5:05

  4. Headcount on any date

    Start dates and leave dates are all it takes to count who was employed on a given day.

    video, 3:50

  5. Length of service in whole months

    Whole months between two dates, worked out the same way in SQL, Python and Excel.

    video, 4:35

  6. A turnover rate that adds up

    Leavers over average headcount for a month, built up one step at a time.

    video, 5:20

everything to watch

Portfolio

  1. Survival curves by department

    Pick a department and see how long its new starters stay, against the whole organisation.

    tool

  2. One question, five languages

    Pick a workforce question and compare the same query written in each language, side by side.

    tool

  3. A headcount and turnover dashboard

    Headcount, starters and leavers month by month, on the synthetic organisation.

    dashboard

everything in the portfolio