Applied AI engineer & educator
I make models intofast, reliablefeatures.
Research tells you what's possible. Shipping tells you what holds up at scale, with real users and real limits. I build the systems that get a model from one to the other, and I teach the engineering behind them.
Now buildingLLM wiki →
Tokenizationtext becomes numbers
Before a model reads your prompt, it cuts the text into tokens. Common words stay whole, rarer ones get split into pieces it has seen before, and each piece is swapped for a number. Those numbers are all the model ever sees.
It matters because you pay, wait and run out of context in tokens, not words.
Tokenization isn’t magic.
Cut wherever the pieces match the tokenizer’s vocabulary
02 Tokens, then 03 Numbers
- Tokentoken id 3404
- izationtoken id 2065
- isntoken id 2125
- ’ttoken id 956
- magictoken id 11204
- .token id 13
Systems I built, and what each one taught me.
Project write-ups are in progress. Each one will split the build into phases and link back to its commits.
Graduate and undergraduate courses, leadership workshops, edX, summer boot camps and Village Games.
- Behavioral analyticsTalk · Stanford University
- Data visualization for business process discoveryTalk · UC Berkeley
- Village GamesNonprofitGame design and programming as a way into STEM.
What I learned from a project or a class, written down while it was fresh.
The first reflections are being written. Each one will link to the project or talk it came from.
Consulting
Bring me in on a hard AI problem.
Scaling up, users who don’t behave the way the plan assumed, and the edge cases that only show up once a model is live.
Start a consulting conversation →Speaking & teaching
Invite me to lecture or teach your team.
Guest lectures, courses, and focused sessions for executive boards and leadership teams.
Ask about a session →