Omar U. Espejel

I’m Omar U. Espejel. I work on AI and crypto at Starknet Foundation. Before that I worked on crypto at StarkWare and machine learning engineering at Hugging Face.

Here I write notes and essays about AI engineering, crypto, proof systems, and product work. A lot of the pieces start as research in progress. I try to make the idea clear enough that someone else can inspect it, reproduce it, or disagree with it.

Downloads
40,239
Models
18
Datasets
8
Followers
85

All-time figures for my Hugging Face repositories, read when this page is built. Downloads are the sum of downloadsAllTime over every public repository, so check the numbers yourself: models, datasets, profile.

Teaching

I teach 4 courses on Platzi, in Spanish: 94 classes and about 14 hours of recorded material, rated 4.8 out of 5 across 472 student reviews.

Course Classes Rating Reviews
Curso de LangChain 37 4.8 192
Curso de Transfer Learning con Hugging Face 17 4.8 114
Curso de Redes Neuronales con PyTorch 24 4.7 93
Curso de Experimentación en Machine Learning con Hugging Face 16 4.8 73

Course figures read from the public Platzi course pages on 2026-08-17.

Podcast

I host Hacia Afuera, a Spanish-language deep tech podcast: 85 episodes and 76 hours of conversation since 2021, with researchers and builders in AI, quantum computing, cryptography and neurotechnology. It is on Spotify, Apple Podcasts and Deezer.

Recent essays

  • Omar U. Espejel 7 min

    The Environment Contract Is Part of the Product

    • agents
    • engineering

    A postmortem on an auth flag regression in an agent-built app, and the deploy contract that stops stale environments from becoming releases.

  • Omar U. Espejel 11 min

    What a Proof Is Allowed to Mean

    • verifiable ai
    • proof systems

    A guide to the missing receipt around proof artifacts: how proof bytes, statement meaning, verifier domain, and replay assumptions become one accepted object.

  • Omar U. Espejel 13 min

    Proof Validity Is Not Statement Validity

    • verifiable ai
    • proof systems

    A proof can verify while the AI claim around it is still relabelable. In a local EZKL-style receipt test, the proof-only path rejected 1 / 7 relabels; the statement-envelope path rejected 7 / 7.

  • Omar U. Espejel 13 min

    Proof Pressure Is Not Just Matrix Multiply

    • proof systems
    • verifiable ai

    The useful transformer-proving question is not whether the model contains arithmetic. It is where proof plumbing gets reused or repeated.

All essays