---
title: 'Research Scientist, Relational Foundation Models at Avra'
canonical: 'https://feeny.ai/job/research-scientist-relational-foundation-models-avra-sao-paulo-hq1dcbsmght9'
type: 'job'
last_seen: '2026-09-12'
---

# Research Scientist, Relational Foundation Models at Avra

- **Company:** Avra
- **Location:** São Paulo, Brazil
- **Employment:** full-time
- **Work type:** remote
- **Posted:** 2026-04-20
- **Last confirmed live:** 2026-09-12
- **Apply:** https://jobs.ashbyhq.com/avra/0ccb2c8e-438d-4e97-bee0-c1ae390a02ac

## Job description

## About Avra

Avra is building relational foundation models for enterprise decision-making in Brazil. Our work focuses on graph-native models for structured, high-stakes prediction problems: credit, fraud, growth, monitoring, and other decisions where entities cannot be understood in isolation. We model companies, people, and the relationships between them as evolving networks, then adapt those representations to customer-specific prediction tasks that plug into existing decisioning systems. We work with internationally recognized research advisors, and we care about research that becomes useful in production.

## The role

This is an applied scientist role with real modeling depth. You will help evolve the thesis, architecture, and applications of Avra’s relational foundation models: how we train them, how we adapt them to specific tasks, and how they generalize across use cases. Day to day, you’ll move between papers, code, experiments, and production constraints. The goal is not to try interesting ideas for their own sake. The goal is to find which ideas improve real downstream models under realistic deployment conditions. We run a weekly research review. Strong papers matter; shipped models matter more.

## What you’ll work on

- New approaches for relational foundation models over heterogeneous and temporal graphs
- GNNs, graph transformers, attention over relations, relative temporal encodings, and other architectures for structured entity networks
- Training objectives such as reconstruction, contrastive learning, generative modeling, supervised learning, and hybrid combinations
- Transfer from foundation representations to downstream tasks through fine-tuning, late fusion, distillation, calibration, and task-specific evaluation
- Rigorous evaluation: temporal validation, leakage checks, ablations, strong baselines, and error analysis
- Large-scale training infrastructure using Ray, including sampling, sharding, memory layout, distributed execution, and throughput optimization
- Performance-sensitive ML systems: data loading, graph sampling, memory efficiency, fused kernels, and training-loop bottlenecks
- Turning research ideas into reliable modeling components used in production

## What we’re looking for

- 5+ years in applied ML research, research engineering, or equivalent high-level ML systems work
- Deep hands-on experience with PyTorch or a similar deep learning framework
- Ability to read current research, identify the core idea, and turn it into a controlled experiment within a week or two
- Experience with graph ML, recommender systems, ranking, time-series models, representation learning, or structured-data domains where strong tabular baselines are hard to beat
- Strong experimental discipline: baselines, ablations, temporal splits, leakage prevention, reproducibility, and honest error analysis
- Comfort with large datasets, distributed training, and the difference between a clean benchmark run and a pipeline that has to work every week
- Engineering judgment to build work that others can maintain
- Clear communication around model behavior, experimental results, and technical tradeoffs

You stand out if

- You have worked with heterogeneous or temporal graphs using PyG, DGL, custom graph tooling, or related systems
- You have used Ray for distributed training, data processing, or serving
- You have written Rust, C++, CUDA, Triton, or fused kernels, or worked seriously with JAX
- You have optimized graph sampling, memory usage, data loading, training loops, or distributed workloads
- You have shipped models into production and monitored how they behaved after deployment
- You have contributed to open-source ML infrastructure, published strong applied research, or built serious internal research systems
- You have worked in environments where the model only matters if it improves a real business metric

## Requirements

- Bachelor’s degree in a quantitative field: Computer Science, Mathematics, Statistics, Physics, Engineering, Economics, or similar
- Master’s or PhD is a plus, not a filter
- Strong written English
- Portuguese is useful, but not required

## What we offer

- Competitive salary, equity, and open compensation bands
- Direct collaboration with founders, research leadership, and experienced AI advisors
- Research budget, paper incentives, and support for publishing when the work is strong and appropriate
- 100% remote work, with a São Paulo office available when you want it
- Flexible time off, national health plan, and extended parental leave
- High ownership over research directions that can become part of Avra’s core platform

If you want to help build foundation models for relational decision-making, not as a benchmark exercise but as infrastructure used by real enterprises on real economic networks, we’d like to meet you.

## About Avra

## Company Overview
- **One-liner**: Avra is a frontier AI lab building a predictive platform for enterprise decisions, powered by a Graph Foundation Model that models the relational economy.
- **Entity Type**: Private (early-stage startup)
- **Headquarters**: São Paulo, Brazil
- **Founded**: 2024
- **Founders**: Bruno Alano (CEO, co-founder) and Viviane Meister (CTO, co-founder)

## Core Business
- **Primary industry**: Enterprise AI / Decision Intelligence (credit, fraud, growth, monitoring)
- **Target customers**: B2B, Enterprise (banks, fintechs, marketplaces, SMB lenders)
- **Mission**: “Model relationships over time. Improve the decision systems enterprises already run.”

## Products & Services
- **Avra Graph Foundation Model**: A pre-trained temporal knowledge graph covering Brazil’s economy – companies, individuals, events, ownership, judicial events, geography. Fine-tuned per customer workspace for credit scoring, fraud detection, and propensity modeling.
- **Avra API & SDK**: Managed endpoints (mTLS) for prediction and explanation, with tenant isolation and audit trails.
- **Avra Playground**: Browser-based environment to experiment with decision flows, replay traffic, and inspect evidence.
- **Enterprise Deployment**: Shadow deployments alongside incumbent models, with batch and real-time inference surfaces.

## Market Standing
- **Valuation / Market Cap**: Not publicly available
- **Key Metric**: Total funding amount not disclosed; backed by “frontier funds across two continents”
- **Notable Investors/Partners**: Not named explicitly, but investors are described as “frontier funds across two continents”. Team alumni include OpenAI, Stone, Itaú, McKinsey, Embraer, XP, HSBC, Santander.
- **Growth Signals**: Small team (~20 people) with 7 open roles; remote-first with HQ in São Paulo; pilot results showing 1.8× NII on Avra-scored cohort, +90% conversion lift, +5.4 p.p. ROC AUC on held-out test set.

## Competitive Advantages
- **Graph-native reasoning**: Models relationships over time rather than flat rows, capturing risk dimensions orthogonal to traditional features (18–22% correlation with existing features).
- **Inductive generalization**: Scores entities never seen before by reasoning through counterparties and graph position.
- **Brazil-native foundation**: Trained on local semantics (CNPJs, corporate groups, informal networks) and legal events.
- **Developer-first, enterprise-ready**: API, SDK, playground, tenant isolation, ISO 27001 in progress, LGPD compliant.

## Strategic Focus
- Expand the Graph Foundation Model platform across more enterprise decision use cases (credit, fraud, growth, monitoring).
- Grow the team (~20 today) with research scientists, data engineers, and deployment strategists.
- Maintain a “small on purpose” approach with deep stack ownership from data ingest to live decision.

## Why Work Here
- **Culture**: “Small team, deep stack, real ownership.” Engineers, scientists, and operators own the work end-to-end.
- **Remote / Hybrid**: Remote-first with hybrid options in São Paulo.
- **Team Background**: Colleagues from OpenAI, Stone, Itaú, McKinsey, Embraer, XP, HSBC, Santander.
- **Perks**: Frontier research that ships into production; publish-grade research with real business impact; no contractors on the model path.
- **Open Roles**: 7 positions including Research Scientist, Senior Data Engineer, Staff Software Engineer, Deployment Strategist, Marketing Lead, Enterprise Sales.

## Sources
1. [avra.ai/about](https://avra.ai/about)
2. [avra.ai/careers](https://avra.ai/careers)
3. [avra.ai](https://avra.ai/?r=0)
4. [avra.ai/en/about](https://avra.ai/en/about)
5. [jobs.ashbyhq.com/avra](https://jobs.ashbyhq.com/avra)

## Other roles at Avra

- [GTM Enterprise - Founding Team](https://feeny.ai/job/gtm-enterprise-founding-team-avra-sao-paulo-55jg60e9y7mm) — São Paulo, Brazil
- [Operator](https://feeny.ai/job/operator-avra-sao-paulo-0gned0b07xbc) — São Paulo, Brazil
- [Deployment Strategist](https://feeny.ai/job/deployment-strategist-avra-sao-paulo-zf85c1zmxq5y) — São Paulo, Brazil
- [Open Application](https://feeny.ai/job/open-application-avra-sao-paulo-t9a0zay6e7qs) — São Paulo, Brazil