Data Science Academy
Data Analyst Certified Professional (DACP)
Issued Jan 2023
Hi, welcome!
I'm
Franklin Ferreira
Staff Data Engineer Data Architecture & Big Data Governance, FinOps & AI
From warehouse to lakehouse, from classic ML to generative AI: I design the data architecture a whole company runs on.
About me Download CV- >50%lower data platform cost
- ~50sources unified in one platform
- 15+people on the team I lead
About
Who I am
In practice, I'm an architect. Before anyone writes a line of code, I think through and decide how the data should be structured, in a warehouse or a lakehouse, in batch or in streaming. Today I technically lead a team of 15+ people (data engineers, analysts and data scientists) spread across Brazil and Romania.
I joined Flutter Brazil to build the foundation. I created the dimensional model and the data lake from scratch, with Python, dbt and a lot of SQL. Then I led the migration of the whole data infrastructure to Databricks: around 50 sources, from Google Analytics 4 to internal systems and third-party tools, with pipelines going from hours to minutes.
As Tech Lead the job got wider. I defined the architectures and ways of working for almost the whole data team, restructured Unity Catalog by domain and led the FinOps work that cut platform cost by more than 50%. As Staff, that scope keeps growing.
I come from the more classic side of machine learning, since college. I once did backpropagation by hand, on paper, before any framework did it for me. I followed the shift to generative AI closely from the moment it appeared, and today I move between both worlds comfortably. I helped build the company's LTV and payback models, and I use generative AI to speed up the day-to-day tools we build.
Before Flutter Brazil, I was a Data Architect at Clicksign. I built the data structure from scratch there, on Redshift, dbt and Airflow, and helped the group grow from 4 to about 20 people. I also spent a few years as a professor of Big Data and Data Engineering, with hundreds of students. Maybe that's why I like explaining things in a way that makes sense to people who don't live in SQL all day.
Profile
Still hands-on every day, in architecture, modeling and platform decisions. When the team needs it, I step in to set priorities and unblock business decisions.
- Role Staff Data Engineer
- Focus Data architecture, Big Data platforms, governance, FinOps, ML and generative AI
- Team 15+ people across Brazil and Romania
- Location Brazil · Remote/Global
Languages
- PortugueseNative or bilingual
- EnglishFull professional proficiency
- SpanishProfessional working proficiency
Skills & technologies
The tools and practices I work with, grouped by what they are for.
Architecture & modeling
Platforms & processing
Transformation, orchestration & cloud
Languages
Data apps & analytics
Practices
AI
Leadership
Impact
What changed
Numbers from my time at Flutter Brazil and Clicksign.
Experience
Where I've worked
Since 2020 I've gone from data scientist to Staff Data Engineer, building platforms, teams and ways of working along the way.
Experience
Career at a glance: pick a role to see the detail.
- Jul 2026 – Present Staff Data Engineer Flutter Brazil Architecture and technical direction of the company's data platform, with a team of 15+ people.
- Jul 2025 – Jun 2026 Data Engineer – Tech Lead Flutter Brazil Platform cost cut by more than 50% and incidents by ~3x; Unity Catalog split into 10+ domains.
- Aug 2024 – Jul 2025 Data Engineer Flutter Brazil Dimensional model and data lake built from scratch; ~50 sources moved to Databricks.
- Oct 2021 – Aug 2024 Data Team Lead Clicksign Data group grown from 4 to ~20 people.
- Jun 2021 – Aug 2024 Data Architect Clicksign 100+ Airflow pipelines, dbt on Redshift, 10+ partner APIs and LGPD data protection.
- Oct 2020 – Jun 2021 Data Scientist (Python Developer) Clicksign Python, PostgreSQL and Airflow integrations for the BI team.
- Jun 2022 – Jul 2024 Adjunct Instructor Let's Code from Ada Big Data and Data Engineering classes for hundreds of students.
Flutter Brazil
Jul 2026 – Present
Staff Data Engineer
A Databricks data platform used by thousands of people across the whole company, with a team of 15+ people (engineering, analytics and data science) across Brazil and Romania.
- Started as Staff Data Engineer, expanding the architecture and technical leadership scope I had been building as Tech Lead.
- Set the architecture and ways of working for a team of 15+ people (10 as Tech Lead), with a wider scope across the whole platform.
- Still hands-on day to day (architecture, modeling and platform decisions). When the team needs it, I set priorities and unblock business decisions.
Technologies: Databricks, Unity Catalog, Delta Lake, Apache Spark, PySpark, Databricks Workflows, dbt, SQL, Python, AWS, Terraform, Docker, Kubernetes, GitHub Actions
Flutter Brazil
Jul 2025 – Jun 2026
Data Engineer – Tech Lead
Led a team of 10 people (engineering, analytics and data science) across Brazil and Romania, with direct and technical leadership, on a Databricks platform of tens of TB.
Achievements
- Cut the data platform's cost by more than 50% with a FinOps effort. The annual bill was in the six figures (USD).
- Reduced incidents and rework by ~3x by rolling out CI/CD with trunk-based development.
- Gave the company instant visibility into issues, with platform-wide observability and near-real-time monitoring of business metrics (latency of a few minutes).
- Restructured Unity Catalog by domain (10+).
Activities
- Defined the architectures and ways of working for almost the whole data team. That went from restructuring Unity Catalog by domain to the ongoing evolution of the company's analytical modeling with dbt, including the CI/CD rollout with trunk-based development.
- Architected the data platform's FinOps effort (clusters, jobs, storage, governance policies), including cost forecasting models so that spend came down in a predictable way.
- Delivered decision-support dashboards, platform-wide observability and near-real-time business metrics monitoring. What the data team built went through my desk.
- Helped create the company's LTV and payback models.
Technologies: Databricks, Unity Catalog, Delta Lake, Apache Spark, PySpark, Databricks Workflows, dbt, SQL, Python, AWS, Terraform, GitHub Actions, Streamlit, Metabase
Flutter Brazil
Aug 2024 – Jul 2025
Data Engineer
Achievements
- Built the company's dimensional model and data lake from scratch. Every area still uses that architecture today.
- Led the migration of the company's entire data infrastructure to Databricks: ~50 sources (Google Analytics 4 among them) in a single platform, with pipelines going from hours to minutes.
Activities
- Dimensional modeling and data lake with Python, dbt and SQL.
- Integrated internal systems, third-party tools and Google Analytics 4 into a single platform of tens of TB.
- Also contributed to the regulatory process that allowed the company to operate legally in Brazil.
Technologies: Python, dbt, SQL, Databricks, Google Analytics 4
Data Team Lead
Achievements
- Grew the data group from 4 to ~20 people across data engineering, data analysis and data quality assurance.
- Directly responsible for the reliability of the data used in company decisions.
Activities
- One of the first engineers on Clicksign's data team, helping build the area from the ground up.
- Led the data engineering, data analysis and data quality assurance teams.
- Built the processes and scaled the team that supported the area's growth.
Technologies: Amazon Redshift, dbt, Apache Airflow, dimensional modeling
Data Architect
Achievements
- Architected practically every data structure the company had in that period, including 100+ Airflow pipelines (modeling and ingestion).
- Led the implementation of dbt on Redshift. It brought quality control, governance, easier modeling, observability and simpler orchestration.
- Built 10+ APIs for sending data to partners and third parties.
- Architected the data protection processes (in compliance with LGPD, Brazil's data protection law, and aligned with GDPR) and the continuity of data delivery.
Technologies: Apache Airflow, dbt, Amazon Redshift, dimensional modeling, data governance
Data Scientist (Python Developer)
- Part of the Business Intelligence team, turning raw data into information and insight through automation and integration.
- Developed scripts and integrations between databases, APIs and REST web services with Python, PostgreSQL and Apache Airflow.
- Contributed predictive analyses and modeling to the BI team.
Technologies: Python, PostgreSQL, Apache Airflow, REST APIs, predictive modeling
Adjunct Instructor
- Taught Big Data and Data Engineering courses to hundreds of students: database modeling, data warehouses, data lakes and lakehouses, programming logic and software architecture, and cloud development.
Topics: Big Data, Data Engineering, data modeling, data warehouses, lakehouses, cloud
Education
- Jul 2022 – Jul 2023 MBA in Data EngineeringXP Educação
- Jun 2021 – Jul 2022 Postgraduate specialization (Lato Sensu) in Data Science and Artificial IntelligencePUCRS · Pontifícia Universidade Católica do Rio Grande do Sul
- 2015 – 2019 Bachelor's degree in Computer ScienceUEZO · Universidade Estadual da Zona Oeste do Rio de Janeiro
Certifications & training
Programs, certifications and courses, most recent first. Each entry links to the original certificate.
Certifications & programs
Data Science Academy
Data Analyst Training
Completed Jan 2023
Data Science Academy
Data Architect Training
Completed Dec 2022
Astronomer
Astronomer Certification: DAG Authoring for Apache Airflow
Issued Dec 2022
Astronomer
Astronomer Certification: Apache Airflow Fundamentals
Issued Nov 2022
Data Science Academy
Data Scientist Certified Professional (DSCP)
Issued Apr 2022
Data Science Academy
Data Scientist Training
Completed Oct 2021
Data Science Academy
Market Intelligence Analyst Training
Completed Jun 2020
Show all 33 courses
- Jan 2023Data Analysis with Microsoft Power BI and BI ClinicData Science Academy
- Jan 2023Cloud Data EngineerXP Educação
- Dec 2022Data Analysis with PythonData Science Academy
- Dec 2022Data Pipeline Design and AutomationData Science Academy
- Dec 2022ETL and ELT Process Architecture and IntegrationData Science Academy
- Dec 2022Modeling Relational and Non-Relational Databases and Data StoresData Science Academy
- Dec 2022LGPD, Data Governance and Metadata ManagementData Science Academy
- Oct 2022Data EngineeringXP Educação
- Jul 2022Data Engineering BootcampHOW
- Mar 2022SQL for Data ScienceData Science Academy
- Feb 2022AWS Cloud Computing Design and AdministrationData Science Academy
- Jan 2022Data Mining with SPSSData Science Academy
- Oct 2021Data Visualization and Dashboard DesignData Science Academy
- Sep 2021Business AnalyticsData Science Academy
- Aug 2021Machine LearningData Science Academy
- Dec 2020Data Engineering with Hadoop and SparkData Science Academy
- Jul 2020Big Data Real-Time Analytics with Python and SparkData Science Academy
- Jun 2020Econometric Analysis in RData Science Academy
- Jun 2020Statistical Analysis and Predictive Modeling of Time Series in PythonData Science Academy
- May 2020Statistical Data Analysis and Machine Learning in RData Science Academy
- Feb 2020Big Data Analytics with R and Microsoft Azure Machine LearningData Science Academy
- Jan 2020Python Fundamentals for Data AnalysisData Science Academy
- Dec 2019R Fundamentals for Data AnalysisData Science Academy
- Oct 2019Linux Operating System FundamentalsData Science Academy
- Sep 2019Financial and Capital MarketsData Science Academy
- Jul 2019Introduction to EconomicsData Science Academy
- Oct 2018Face Recognition with Python and OpenCVUdemy
- Oct 2018Face and Object Recognition with Python and DlibUdemy
- Oct 2018Face Detection with Python and OpenCVUdemy
- Jan 2018Web DeveloperUdemy
- Jan 2018E-commerce Integration with PHP 7 and PagSeguroUdemy
- Jan 2018Mastering MySQL DatabasesUdemy
- Jan 2018PHP 7Udemy
Talks & research
Talk: Airflow — who said you can't process data in it?
Airflow is a platform to programmatically author, schedule and monitor workflows. I asked whether it could also be used to process data: what the benefits would be, how to build consistent, simple pipelines for low data volumes, and which pipeline architectures are possible.
Data Science Academy Podcast
I talked about my career path, mindset, joys and frustrations, how I landed my first job as a data scientist and what the day-to-day looked like.
UEZO
Aug 2018 – Aug 2019
Undergraduate research
I worked on understanding and refining recognition and detection techniques for objects like lines, human faces and cells. As a branch of that research, I developed an edge-detection optimization analysis for cells, aimed at making it easier to build 3D models in biology, pharmacology and medicine.
Work
Selected work
Case studies from the platforms and teams I've led, followed by earlier machine learning projects.
Databricks migration and lakehouse foundation
~50sources unified in one platform
Starting point~50 sources, from Google Analytics 4 to internal systems and third-party tools, on a platform of tens of TB. Thousands of tables and thousands of jobs a day.
What I didMoved the whole data infrastructure to Databricks and built the company's dimensional model and data lake from scratch, with Python, dbt and a lot of SQL.
What changedPipelines went from hours to minutes, and the dimensional model and data lake are still what every area works with today.
- Databricks
- dbt
- Python
- SQL
- Google Analytics 4
FinOps and cost forecasting
>50%lower data platform cost
Starting pointAn annual data platform bill in the six figures (USD).
What I didArchitected the FinOps effort across clusters, jobs, storage and governance policies, with cost forecasting models so that spend could come down in a predictable way.
What changedMore than 50% off the data platform's cost.
- Databricks
- FinOps
A data infrastructure from scratch at Clicksign
100+Airflow pipelines
What I didArchitected nearly every data structure the company had in that period, with 100+ Airflow pipelines for modeling and ingestion, and led the implementation of dbt on Redshift.
What changedQuality control, governance, easier modeling, observability and simpler orchestration.
- Apache Airflow
- dbt
- Amazon Redshift
- dimensional modeling
Unity Catalog by domain
10+Unity Catalog domains
What I didRestructured Unity Catalog by domain and looked after data modeling and quality, so provisioning stays reliable and secure.
What changed10+ domains, with architectures and ways of working defined for almost the whole data team.
- Databricks
- Unity Catalog
- dbt
Observability and near-real-time metrics
What I didBuilt apps for platform observability and for business metrics monitoring, all near-real-time, along with decision-support dashboards.
What changedInstant visibility into issues, with a latency of a few minutes.
- Streamlit
- Metabase
CI/CD with trunk-based development
~3xfewer incidents and less rework
What I didRolled out CI/CD with trunk-based development and helped steer how the company's analytical modeling evolves with dbt.
What changedIncidents and rework dropped by ~3x.
- dbt
- CI/CD
- trunk-based development
Data protection and partner APIs
10+APIs for partners and third parties
What I didArchitected the data protection processes (in compliance with LGPD and aligned with GDPR) and the continuity of data delivery, and built 10+ APIs to send data to partners and third parties.
- LGPD
- GDPR
- APIs
- data governance
LTV and payback models
Starting pointQuestions the models answer: acquisition spend, return per customer and payback time, the most valuable segments, and whether campaigns and promotions paid off.
What I didHelped build the LTV and payback models (SQL and dbt, with statistical and machine learning modeling, refreshed by pipelines and shown in dashboards). My part: the input data and architecture, the pipelines and platform, and validating results with data science and the business.
What changedMarketing got a numeric basis to decide where to invest.
- dbt
- SQL
- Machine learning
Generative AI for day-to-day tools
Where I use itCode in SQL, dbt and Python, documentation and data quality tests, natural-language assistants and queries, and data analysis.
What I didI use generative AI to speed up the day-to-day tools we build, from prototypes in Lovable to dashboards in Streamlit and Metabase. I introduced it to the team, set the guidelines for using it and built AI solutions for the company.
What changedPrototypes arrive faster, the team spends less time on repetitive work and more people build tools.
- Cursor
- OpenAI
- Anthropic
- DeepSeek
- GLM
- Lovable
- Streamlit
- Metabase
Decisions and trade-offs
The reasoning behind three platform decisions: what weighed, the trade-offs I managed and what I would change.
Why Databricks, and what I would change
What weighed in the choiceA single platform for batch processing, ML and BI; governance with Unity Catalog; cost and performance; and the skills of the team.
What else was on the tableStaying on Redshift and the native AWS stack, Snowflake, and BigQuery on Google Cloud.
Biggest risks in the migrationMigrating without stopping the business, reaching data parity, and the cost during the transition.
What I would do differentlyStart FinOps earlier, and define domains and owners earlier.
- Databricks
- Unity Catalog
- FinOps
FinOps: priorities and trade-offs
What I prioritizedVisibility of spend, the biggest cost drivers, cluster sizing, and job scheduling.
Trade-offs I managedCost against data freshness, cost against performance, autonomy against rules, and engineering time against savings.
What kept the savings in placeCost forecasting models, cost attributed by domain or owner, spend alerts and dashboards, and policies and guardrails.
What I would do differentlyTag and attribute cost by domain and owner from day one, set up automatic spend alerts earlier, and involve the domains earlier.
- Databricks
- FinOps
Unity Catalog by domain: reasons and trade-offs
Why by domainClear ownership by area, access and security by domain, cost attributed by domain, and discovery and data quality.
Trade-offs I managedCentral control against team autonomy, how many domains (granularity), rework of tables and permissions, and standards against flexibility.
What worked in practiceDomain owners and agreed standards, and provisioning as code.
What I would do differentlyAutomate provisioning and permissions earlier, and set up lineage and documentation for the domains from the start.
- Databricks
- Unity Catalog
Earlier machine learning projects
Analyses built while studying data science, kept here for reference.
Show the 13 projects
National Institute of Diabetes and Digestive and Kidney Diseases
In this project we use Python to predict, based on diagnostic measurements, whether a patient has diabetes.
Big Data Analytics & Machine Learning
Customer Churn in Telecom Operators
In this project we use Python to build a machine learning model that predicts whether a customer may cancel their plan, and the probability of that happening.
Big Data Analytics & Machine Learning
Santander Customer Satisfaction
In this project we use Python to build a machine learning model that predicts whether a customer is satisfied with their banking experience.
Big Data Analytics & Machine Learning
Netflix Investor: Quarterly Earnings
In this project we use R to prepare a statistical summary of Netflix's data, showing the company's financial health and other relevant information that can help potential investors decide whether to invest.
Big Data Analytics & Forecasting
Warm Up: Predict Blood Donations
In this machine learning project we use R to predict whether a blood donor will donate within a given time window. The goal is to predict the values of the last column (Made Donation in March 2007) to determine whether the donor gave blood in March 2007.
Big Data Analytics & Machine Learning
TalkingData AdTracking Fraud Detection Challenge
In this project we build a machine learning model to determine whether a click on an online ad is fraudulent.
Big Data Analytics & Machine Learning
Grupo Bimbo Inventory Demand
In this machine learning project we develop a model to accurately predict inventory demand from historical sales data. This keeps shoppers from finding empty shelves for any of the more than 100 bakery products and reduces the amount spent on refunds to store owners for surplus products unfit for sale.
Big Data Analytics & Machine Learning
Warm Up: Machine Learning with a Heart
Heart disease is the leading cause of death worldwide. To learn how to prevent it, we must first learn to detect it reliably. In this project we build predictive models to classify whether a patient has heart disease.
Big Data Analytics & Machine Learning
Boston Houses
In this project we build models to predict house prices in Boston.
Big Data Analytics & Machine Learning
Big Data Analytics on Twitter - Part I
The goal of this analysis is to explore different techniques and tools to capture, manipulate and transform data from the social network Twitter. The analysis seeks to understand the sentiment each tweet conveys, enabling the extraction of information, knowledge and wisdom. This technique helps decision makers analyze the sentiment of their target audience toward a given topic, for example to determine whether a marketing campaign is received positively, negatively or neutrally.
Big Data Analytics
Big Data Analytics on Twitter - Part II
The goal of this analysis is to explore different techniques and tools to capture, manipulate and transform data from Twitter. We introduce the comparison of user sentiment toward two topics (two keywords). This technique helps decision makers understand the sentiment of their target audience toward multiple topics, for example to determine whether one marketing campaign was received more positively or negatively than another.
Big Data Analytics
Big Data Analytics on Twitter - Part III
The goal of this analysis is to explore different techniques and tools to capture, manipulate and transform data from Twitter. We classify the sentiment each tweet conveys and determine its polarity. This technique helps decision makers understand the sentiment of their target audience toward a given topic, for example to determine whether a marketing campaign caused surprise, anger, fear, disgust, joy, etc.
Big Data Analytics
Big Data Analytics on Twitter - Part IV
The goal of this analysis is to explore different techniques and tools to capture, manipulate and transform data from Twitter. We evaluate how frequently a given keyword is used in a geographic region. This technique helps decision makers understand the sentiment of their target audience toward a given topic, for example to determine in which cities a marketing campaign was most talked about.
Big Data Analytics
How I see it
In the end, what matters is whether the data became a decision.
Expertise
Where I can help
If you're building or fixing a data platform, these are the areas where I can help.
Data architecture & modeling
Working out how the data should be structured before any code exists: dimensional models, warehouses, data lakes and lakehouse foundations.
Data platforms: design & migration
Designing data platforms and bringing many sources into one: warehouses with dbt and Airflow, lakehouses, and a ~50-source migration that took pipelines from hours to minutes.
FinOps & cost forecasting
Clusters, jobs, storage and governance policies, with forecasting models so the savings are predictable.
Governance, quality & LGPD
Governance and data catalogs (AWS Glue Data Catalog, Unity Catalog, Hive Metastore and open-source catalogs), data quality, and data protection processes that follow LGPD and GDPR.
Observability & metrics
Seeing platform issues and business metrics in near-real-time, with a latency of a few minutes.
Engineering practices
Pipelines and orchestration with Airflow and dbt, CI/CD with trunk-based development, and provisioning that stays reliable and secure.
Technical leadership
Leading data engineers, analysts and data scientists: setting architectures, ways of working and priorities.
Generative AI for data teams
Bringing generative AI into a technical team's daily work so that building tools and prototypes gets faster.
Contact
Let's talk
I'm always up for talking about data architecture, generative AI applied to a technical team's day to day, or the data market in Brazil.
Opportunities
Open to any kind of opportunity: full-time, contract or partnership.
Talk about an opportunityProjects
Architecture reviews, platform migrations, FinOps, data governance or any other data challenge.
Talk about a projectBrazil · Remote/Global