Sebastián Sarmiento

Learning scientist · Curriculum engineer — Santiago, Chile

Back to
learning in
the world.

We use mathematics and AI to read the real systems of the planet and turn them into living learning, not isolated exercises. I work at the seam where psychometrics, data mining, curriculum, and product meet.

Read the manifesto
Coastal campus connected to city, fields, and renewable energy by a mathematical learning network
The global learning node

No one designed the school.
Everyone pushed it.

The school as we know it was not designed: it was the point where the State, the factory, the church, and the empire converged. Two hundred years later, we still teach inside that form — and in no discipline does it show more than in mathematics.

AI can take over the procedural — the algorithm, the technique, the practice that builds fluency — until mastery on any standardized test takes the least time possible. Let the test stop being the center and become a formality: let us break it by saturation, with everyone answering at a level where it can no longer stratify anyone.

And then the classroom is free for what no machine can answer: understanding why, and imagining solutions to water scarcity.

I can't do this alone.

Read the full manifesto

Work

Full index

A mastery model that survived a real curriculum

How we replaced a checklist of standards with a mastery model that teachers could act on — and that statistics could defend.

Framework · Measurement Middle-school mathematics 9 months — 2023

Knowledge tracing without the black box

Bayesian Knowledge Tracing made legible to the people who teach and the people who build — not just to the model.

Psychometrics · EDM Adaptive mathematics practice 6 months — 2022

An item-validation pipeline teachers trust

A gate that lets only well-evidenced items into the live bank — and tells you, in plain terms, why an item was rejected.

Assessment Assessment & item banking — confidential Ongoing — 2020

In production: a bank of 700+ verified items generated with an AI pipeline and automated quality gates · 19 studies classified by ESSA evidence levels (I–IV) · item design and validation for the national university admission system · IB MYP mathematics leadership · work presented at ISTELive 25.

Essays

Full archive

Translating IRT for people who build products

Item Response Theory is not a reporting feature — it’s a way of deciding what a score is allowed to mean. Here’s the working translation for product teams.

Essay Psychometrics & EDM 2026

Bayesian Knowledge Tracing, honestly

BKT is four numbers and a strong assumption. Used honestly it’s a discipline for accumulating belief; used carelessly it launders guessing into certainty.

Research Note Psychometrics & EDM 2026

Topics

The intellectual index — 5 threads

  1. Mastery Measurement

    What a “mastered” cell is allowed to mean — and the evidence that licenses the claim.

  2. Learning Science → Product

    Translating measurement and learning science into product decisions that survive a classroom.

  3. Psychometrics & EDM

    Item Response Theory, Bayesian Knowledge Tracing, and educational data mining, kept honest.

  4. Math Education

    Curriculum, assessment design, and the unit as an argument rather than a checklist.

  5. AI in Curriculum

    Where AI serves the argument in a unit instead of the novelty — including work shown at ISTELive 25.