Hordago Labs
Deterministic, auditable, Claude-assisted. Built in public while I finish my PhD at UCSD / Salk.
Jeff Jaureguy
Founder · PhD candidate, UCSD / Salk
Why Hordago
hor dago · eu.
My aitaxtxi and amatxi — my grandparents, Martin Jaureguy and Mary Changala — taught me mus, the Basque card game full of bluffs, signals, and shouting hordago! when you're all in.
Hordago comes from the Basque hor dago — literally, there it is. At a mus table, it's the call you make when you stake the entire match on the cards in your hand. Not a bluff. A declaration that the evidence is on the table.
A good mus player throws a calculated hordago: they read the opponent, they commit, and they win or lose on what's verifiable. Not on rhetoric.
Hordago Labs is the same bet, applied to AI for biology. We don't generate
science — we compile it. Every figure ships with a manifest.json
you can rebuild bit-identical. The argument is never the paper — it's the
artifacts you can re-derive.
"Life is like mus. I'm all in. Hordago!"
— Jeff Jaureguy · Basque American essays
About the founder
I'm a PhD candidate at UC San Diego, based at the Salk Institute (McVicker Lab), building deep learning models that predict how rare and noncoding variants perturb gene regulation in specific immune cell types — with a focus on type 1 diabetes and autoimmune disease.
That research shaped how I build Hordago: compile the artifacts, ground every claim, keep humans in the loop on every critical decision. Scientific AI can already generate fluent papers faster than any reviewer can verify them — Hordago exists to close that gap.
Outside the lab I sit zazen, train yoga, box, and run a 60-member Discord tennis league. Cognition is fragile. Confidence is cheap. Rigor is what keeps you honest.
Awards & honors
What this is
It's a platform for evidence-first AI in biology — a thesis, a set of open tools, and a small consulting practice. I'm still in grad school; Hordago is what I do with the rest of my time. Pre-VC. No fundraising. Built in public.
28
Domain Operating Systems
Thin shells, one per scientific domain — genomics, CRISPR, drug discovery, clinical, structural biology.
5
Reasoning plugins
co-writer · co-scientist · sophia-claudette · paper-intelligence · knowledge-graph
2,064
biocontext7 skills
SHA-256-verified MCP data layer. Served to Claude Code, Cursor, VS Code, JetBrains.
✓
Evidence Guardian
Grounding + provenance + contradiction sweeps. Human gates on every critical decision.
Work with me
I'm taking on a small number of engagements while finishing my PhD. Bias is toward work that makes evidence-first AI for biology better, not bigger.
01 · Retainer or project
Biotech, pharma, diagnostics, and AI-infrastructure teams building evidence-first AI for biology. Retainer or scoped project — architecture review, plugin build, provenance design, or technical advisory.
jeff@hordagolabs.com →02 · Advisory
Select advisory engagements for biotech and pharma teams building evidence-first AI. Architecture review, plugin design, provenance gates. Every engagement ships as a Hordago manifest.
jeff@hordagolabs.com →03 · Public
biocontext7, Mus as a metaphor, evidence-first AI, reproducibility in the LLM era. Blog on jeffjaureguy.com, papers on Google Scholar, plugins on GitHub.
Invite me →