CopaData
Checks what fans say about the World Cup against decades of match data, with the verdict rule set before looking.
- Python
- pandas
- Astro
Data & AI Engineer · Porto Alegre, Brazil
Eduardo Stuani
I build data pipelines and LLM systems that survive contact with real data, at scale and under cost constraints.
About
Data Engineer at Econodata, working where large-scale data pipelines meet applied AI. I build the systems that turn public and web data into intelligence on millions of Brazilian companies, from ETL on Airflow and Kubernetes to LLM agents in production.
I care about what holds up with real data: pipelines that run faster, cost less and fail loudly instead of silently, and models that earn their cost. That’s why I measure before I choose, as in the benchmarks below.
- AI agents
- LLMOps
- RAG
- LLM evaluation
- Information retrieval
- Data infrastructure
Work
Fipe
A dashboard to explore 20 years of Brazilian car prices: rankings, filters and price history by model or plate.
- Python
- pandas
- Shiny for Python
Enem
A dashboard of my own ENEM practice exams: score trends by subject and answers I knew versus guessed.
- Python
- pandas
- Shiny for Python
Benchmarks
Laya vs LLMs vs BERT
Laya, BERT models and LLMs tested side by side on routing consumer complaints and flagging fraud.
- Python
- Hugging Face Transformers
- OpenAI API
Experience
Econodata
B2B SaaS platform with data and intelligence on millions of Brazilian companies.
Full-time · Remote
Data Engineer Jun 2026 – now
- Designed, evaluated and deployed LLM-based agents for large-scale company data inference, choosing production configurations on a cost-versus-accuracy basis.
- Delivered LLM-generated data to production at scale, with dual writes to relational and search databases, auditability and safeguards for existing data sources.
- Improved and validated production machine learning models serving data for millions of companies.
- Established evaluation and governance practices for AI experimentation; reported analyses and technical recommendations to the CEO.
Junior Data Engineer Jan 2026 – May 2026
- Optimized heavy ETL workloads on Airflow and Kubernetes, redesigning processing strategies to cut runtime and memory footprint.
- Built monitoring and cost alerting for LLM API usage across providers.
- Scaled and stabilized the AI services running on top of the vector database, fixing retrieval issues in production.
Data Engineer Intern Aug 2024 – Jan 2026
- Built from scratch a natural-language search microservice in FastAPI with hybrid search (Elasticsearch + vectors), integrated into the platform’s main page.
- Implemented monitoring and alerting with Grafana and Prometheus across VMs, Kubernetes clusters and microservices.
- Built and maintained web crawlers and ETL pipelines over structured and unstructured data.
Petrobras
Brazil’s largest oil company.
Research scholarship · Geológi-QA project
Undergraduate Researcher Dec 2023 – Aug 2024
- Built a RAG-based question-answering tool and improved information retrieval over structured and unstructured energy-sector data.
CNPq
Brazil’s National Council for Scientific and Technological Development.
Research scholarship
Undergraduate Researcher Sep 2022 – Jan 2023
- Collected and analyzed Twitter data on Brazil’s 2022 elections, using data mining and network analysis in Python.
Education
BSc in Computer Science
Universidade Federal do Rio Grande do Sul (UFRGS) · 2022 – 2026
Certifications
Preparing for Google Cloud Certification: Cloud Engineer
Google Cloud · 2024
Toolbox
AI
- LLMs
- AI agents
- Tool calling
- MCP
- RAG
- Embeddings
- LLM evals
- LangChain
- LangSmith
- Fine-tuning
Data
- Python
- SQL
- Apache Airflow
- PostgreSQL
- Elasticsearch
- Milvus
- Neo4j
- ETL/ELT
Platform
- Google Cloud
- Kubernetes (GKE)
- Docker
- FastAPI
- CI/CD
- Git
Observability
- Grafana
- Prometheus
- Cost alerting
Contact
Want to talk about data, LLMs or anything on this site? The fastest way to reach me is email or LinkedIn.