Head of Research

Foundation models learned to read text, images, and code. The data that runs the economy hasn't had its moment yet. It's structured: customers, accounts, transactions, and claims, stored in relational databases, connected as graphs, and moving through time. Every credit approval, fraud block, and retention offer is a prediction over that data. Today each one is still a bespoke model, rebuilt by hand for every task, schema, and company.

We're building foundation models for structured data, such as relational databases, graphs, and time series. Take a relational database: one pretrained model should read any company's database, understand its schema from the names of its tables and columns, and adapt to a new prediction task without a feature-engineering project or a model trained from scratch.

Avra exists to build those models and the platform that turns their predictions into decisions running in production, in our cloud or inside an enterprise's own infrastructure. We're an AI research lab with our home in São Paulo, backed by Sequoia and Andreessen Horowitz, and we ship with enterprise design partners in banking, payments, telecom, and commerce, including Mercado Livre, Inter, PicPay, Banco BV, and Vivo.

We are a small group of researchers, engineers, and company builders, with our home in São Paulo and a team spread across San Francisco, Latin America, and Europe. Research is one team working on one portfolio: pretraining, post-training, data, and evaluation for our foundation models. Research leadership is shared between two peers. Our Chief Scientist owns scientific direction and the standard of evidence. The Head of Research owns whether the team delivers.

Our researchers have published widely at venues like NeurIPS and ICLR. You can find our work on our research page.

Make Avra's relational foundation model the best in the world, and build the team that keeps it there.

• The portfolio. What Research works on this cycle, what it stops, and who owns each initiative. You set this together with the Chief Scientist and are accountable for the outcome.

• The team. Hiring, developing, and allocating researchers, and making Avra the place where the best researchers in relational and tabular ML want to work.

• The cadence. Written hypotheses, research reviews, decision records, and an experiment ladder that runs small and cheap before it scales. Plus a compute budget that's spent on purpose.

• The interfaces. Research hands candidate models and their evidence to ML Systems for governed release. It works with our data teams on pretraining corpora and supports Forward Deployment when a customer's model needs adapting. You make those contracts explicit and keep them working.

• The commitments. What our models can promise a customer, by when, and on what evidence. You'll explain this clearly to executives, to customers' risk teams, and to our own commercial team.

You'll stay hands-on: reading code, running experiments, and going into the details of a training curve when something looks wrong.

• You've led a research or applied ML team whose models ran in production and moved a real metric

• Deep technical credibility in modern deep learning, especially transformers, pretraining, and post-training

• Portfolio judgment: you know when to scale a promising idea and when to kill it, and you've done both

• You've hired and grown researchers who went on to do the best work of their careers

• Operating rigor: reproducibility, evaluation protocols, and compute budgets

• Clear written and spoken English

• Pretrained foundation models at scale

• Worked on tabular or relational foundation models, in-context learning, or meta-learning

• Built models for credit, fraud, or other regulated decisions

• Portuguese, since our design partners are in Brazil

• A record of research or open-source work that other researchers build on

1. Intro call with the CTO (45 min). Your path, and what you'd want this role to be.

2. Deep dive (60 min). A team or research program you led: the bets, the results, and the calls you'd make differently.

3. Portfolio exercise (90 min). We share context on where the model is today. You come back with a written plan for the next two quarters and walk the CTO and Chief Scientist through it.

4. Research team (60 min). A session with the researchers you'd lead, without the founders in the room.

5. Founders (45 min).

6. References.

• Location: Remote. We hire in most countries through Deel. The team works across San Francisco, Brazil, Uruguay, and Europe, and we get together in person regularly, in São Paulo and elsewhere.

• Compensation: Competitive salary and meaningful equity.

• Compute: A compute and research budget that you'll own.

• Benefits: Flexible time off, health coverage, and extended parental leave.