AI applied to critical workflows
I have experience with audit automation, contract reading, clause extraction, regulatory compliance and technical information retrieval using LLMs and multi-agent systems.
NLP
MLOps
LLMOps
GenAI
Recruiter Introduction
I turn data, automation and AI into clear, practical and scalable solutions.
I am Flávia, a Senior Data Scientist with 6 years of experience in AI, data and automation. My work combines Generative AI, NLP, machine learning and the full data process that supports data science, from structure and data quality to modeling, automation and real-world application.
- Experience
- 6 years in data, AI and analytics
- Role
- Senior Data Scientist
- Specialty
- Generative AI, RAG and data architecture
- Academic Background
- UnB, USP e IESB
Executive Summary
What I bring to data and product teams
Strong science and engineering foundation
I work with Python, Spark, Databricks, Delta Lake, MLflow, Streamlit, LangChain and governance-oriented pipelines designed for monitoring and production.
Communication and continuous learning
Alongside hands-on delivery, I keep strong technical content, recent Data Engineering certifications and academic research in Data Science.
Professional Journey
Experience built across Industry 4.0, government, consulting and applied AI
FNDE / G4F ecosystem
Role: Senior Data Scientist
Structured an analytics environment in Azure and Databricks with Medallion Architecture, scalable pipelines, governance and a strong foundation for audit and intelligence use cases.
BBTS
Role: Senior Data Scientist / Senior AI Engineer
Built an intelligent payment calendar system from PDF contracts using LangChain, CrewAI, Databricks Workflows and Delta Lake.
Compass UOL
Role: AI-focused Data Scientist
Created LLM-based solutions for technical PDF extraction, regulatory RAG, Streamlit validation flows, MLflow versioning and autonomous agents.
EVCOMX · service provided to Petrobras
Role: Data Scientist / AI Specialist
Worked on machine learning, NLP, technical data extraction, Streamlit, few-shot prompting and the evolution from regex pipelines to LLM-based solutions.
MDS · Ministry of Social Development, Family and Fight Against Hunger
Role: Data Analyst
Analyzed large government datasets such as Bolsa Família, BPC and Cadastro Único using Big Data, SQL, Python scripts and analytical dashboards for decision support.
Presidency of the Republic
Role: Data Science Intern
Automated news collection with web scraping, used NLP with spaCy for NER and summarization, built Power BI visualizations and explored experimental deep learning projects.
Areas of Practice
A hybrid profile combining technical depth, real-world execution and business context
Generative AI and LLMOps
Development of solutions with LLMs, RAG, agents and structured workflows for extraction, analysis, automation and decision support in corporate contexts.
Machine Learning and NLP
Experience with classification, supervised models, natural language processing, embeddings, deep learning and applications focused on text, documents and unstructured data.
Data, Architecture and Pipelines
Design of analytics environments, scalable pipelines, governance, data quality and modern architecture to support analytics, machine learning and AI.
Analytics and Decision Support
Exploratory analysis, dataset reconciliation, dashboards, monitoring and translation of data into clear insights for technical, operational and business teams.
Education and Credentials
Credentials that support the hands-on side of my work
UnB, USP e IESB
Master's in Applied Computing at UnB, MBA in Artificial Intelligence and Big Data at USP, and a Bachelor's in Data Science and Artificial Intelligence at IESB.
Data Science and legal texts
Data Science research line focused on named entity recognition in legal texts, supported by a strong base in databases, large-scale data mining and applied experimentation.
Data Engineering e Databricks
Recent certifications in Airflow, Spark, Snowflake, BigQuery, Modern Data Stack and Databricks tracks connected to data preparation for machine learning and retrieval agents.
Highlighted Credentials
Certificações
Databricks
Badge tied to machine learning tracks and modern data ecosystems.
Retrieval Agents
Building retrieval agents and applied workflows within the Databricks ecosystem.
Palantir Foundry
Training focused on building data solutions and operations on enterprise platforms.
AIP Foundations
Complementary foundation in AI applications and workflows in the Palantir ecosystem.
Data Engineering
Data Architect 4.0 training with 360 hours, plus focused certifications in Spark, Airflow, BigQuery, Snowflake and Data Engineering fundamentals.
Databricks and Retrieval
Tracks completed in March 2026 in Data Preparation for Machine Learning and Building Retrieval Agents on Databricks.
AI and Agents
Certification in Introduction to LangGraph, an academic talk on ChatGPT at USP and additional tracks focused on applied AI solutions.
Governance and Regulated Environments
Petrobras courses in LGPD, information security, information classification and handling, human rights, conflict of interest and anti-discrimination.
Academic Experience
Read my thesis projects connecting research, prototyping and product thinking
NER in legal texts with spaCy and a published application
In my undergraduate capstone, I developed two spaCy models fine-tuned in Brazilian Portuguese for named entity recognition in the legal domain, using the LeNER-Br dataset and publishing a working application on Hugging Face.
- Theme: extraction of legislation, jurisprudence, people, time, location and organization entities from legal texts.
- Results: F1 of 81.42% in the small model and 83.76% in the large model, both above 80%.
- Pitch: applied research in legal NLP, comparative evaluation and a usable end-user deliverable.
CNN for COVID-19 support in chest X-ray diagnosis
In my MBA thesis, I developed and validated a convolutional neural network model to classify chest X-ray images, focused on supporting COVID-19 diagnosis.
- Dataset created with 2,089 images, including 505 COVID-19 cases and 1,580 normal cases.
- Results: 98.40% test accuracy, with hyperparameter tuning and data augmentation.
- Pitch: deep learning applied to healthcare, experimental design and public code on GitHub.
Digital Presence
Where to explore my work further
LinkedIn
Posts, articles and professional positioning.
GitHub
Code, experiments and technical paths connected to my work.
Kaggle
Notebooks, challenges and applied data science exercises.
Hugging Face
AI exploration, models and experiments connected to the open-source ecosystem.
Medium
Authorial content on graphs, Spark, Python, statistics and AI libraries.
Current Linktree
Transition point until this website becomes the main destination.
Contact
Get in touch and let’s talk
I am always open to new opportunities.