Breaking: Teramot connects any AI model to your real data in under an hour Make AI understand your business Formerly Halley - now AI-native data infrastructure Claude + ChatGPT + Gemini, meet your database The Gold Table: clean, query-ready, auditable Founded 2022 - San Francisco x Cordoba Seed-backed by Natan Ventures, CITES & more No data engineers required
Teramot logo
EXHIBIT A. The mark of a company that wants your AI to stop guessing. Teramot - the layer between the chatbot and the database it was never allowed to touch.
Company File / AI Data Infrastructure

Teramot.

"Make AI understand your business."

A swarm of AI agents that reads your databases, fixes the messy parts, and hands your favorite model a clean answer - in under an hour, no data team required.

The Scene

Everyone's AI demo works. Until it meets your data.

The slide deck is glorious. The chatbot answers in full sentences, cites a figure, even cracks a small joke. Then someone in the back row asks the question every executive eventually asks: "Can it do that with our numbers?" The room goes quiet. Because the company's real data does not live in a tidy demo. It lives in three databases nobody fully trusts, a CRM two acquisitions deep, an ERP that predates the cloud, and a spreadsheet a finance lead guards like a family heirloom. This is the gap Teramot was built to close - the unglamorous space between a brilliant model and the messy systems a business actually runs on.

Teramot's claim is refreshingly small. Not artificial general intelligence. Not a chatbot that replaces your staff. Just this: connect any AI model - Claude, ChatGPT or Gemini - to your company's real data, and do it before the coffee gets cold. The execution behind that small promise is the interesting part.

Connect any AI model to your company's real data. In under an hour. No technical team needed. - Teramot, company positioning
2022
Founded (as Halley)
~26
Team Members
<1hr
Time To Connect
3
Models Supported
The Machine

What actually happens under the hood

Teramot is not a single bot. It is a system of specialized AI agents that treat your data the way a careful analyst would - by reading it first, doubting it second, and only then transforming it. The output has a deliberately humble name: the Gold Table, a clean, production-ready layer any AI assistant can query on day one.

1

Connect

Plug into existing databases, warehouses, ERPs, CRMs and APIs - no rip-and-replace.

2

Inspect

Agents analyze the schema and flag data-quality issues before they poison an answer.

3

Transform

Propose transformations and generate transparent, auditable SQL - reviewable before deploy.

4

Query

The Gold Table is immediately queryable from Claude, ChatGPT or Gemini via MCP.

The Goods

Three pieces, one promise

AI Data Infrastructure

The platform: specialized agents that build production-ready data pipelines without data engineers or a six-month implementation.

The Gold Table

A structured, query-ready data layer the agents assemble - the clean source of truth your AI assistant talks to.

Teramot MCP

A Model Context Protocol layer that lets any AI model query and use your data anywhere, with humans kept in the loop on every transformation.

The Stakes

Transparent SQL, not a black box

Most "ask-your-data" tools win the demo and lose the audit. Teramot's bet is the opposite: every transformation is visible, every query is reviewable, and a human can veto the SQL before it ships. In a world nervous about what AI does with sensitive numbers, that auditability is less a feature than a permission slip.

What you can do with it

  • Ask business questions in plain language and get answers from real data
  • Stand up pipelines without hiring a data-engineering team
  • Review and approve the SQL before anything touches production
  • Bring your own model - Claude, ChatGPT or Gemini

Who it's for

  • Mid-market companies the AI wave skipped over
  • Analysts and ops teams drowning in integration work
  • Leaders who want decisions one sentence away
  • Enterprises wary of ungoverned AI access to data
The People

Built by people who were doing AI before it was a hashtag

Founder and CEO Bruno Ruyu says he has been working in AI since 2005 - a detail that lands differently in a field where most resumes start in 2023. He brings a background in data analytics and financial-risk modeling, a stint at Stanford's Graduate School of Business, and a co-founder, Gabriel Puertas, running operations.

BR

Bruno Ruyu

Founder & CEO
GP

Gabriel Puertas

Co-Founder & COO
JC

Jon Castor

Independent Director
FC

Federico Ceccatto

Technical Leader
The Money & The Milestones

A cross-border seed story

2022

Founded as Halley

The company starts with a question - can AI reliably access company data through conversation?

2022 - Seed

Seed funding

Backed by Natan Ventures, CITES, Ryta Zasiekina, Newchip, Startup Ole Accelerator and MANA Tech. Reported totals range up to ~$600K across sources.

2024

Jon Castor joins the board

A product-strategy veteran signs on as independent director.

2024+

Rebrand to Teramot

The company sharpens around "Make AI Understand Your Business," shipping the Gold Table and MCP-based teramot-dev layer.

Marginalia

Things you can drop at a dinner party

Origin

Teramot was originally called Halley - the rebrand came with the pivot to AI-native data infrastructure.

The Founder

Bruno Ruyu has been doing AI since 2005 and studied at Stanford GSB.

Two Homes

A San Francisco address with deep engineering roots in Cordoba, Argentina.

The Name

The flagship output is humbly called a "Gold Table" - a clean, query-ready data layer.

The Return

Back to that quiet room

Replay the meeting. The chatbot is up, the slide is glowing, and the question comes again from the back row: "Can it do that with our numbers?" Only this time the data is already connected. The agents have read the schema, flagged the duplicates, written the SQL someone actually approved. The answer arrives in plain language, sourced from the company's own Gold Table - and the person who asked can see exactly where it came from. The gap between a good demo and a real answer is the whole game. Teramot's wager is that closing it quietly, in under an hour, is worth more than any one model's cleverness. The room, for once, does not go silent.

Connect

Find Teramot

► Watch product demos & interviews on YouTube