About me

My name is Danillo Barros

I have major in statistics. I studied in a public college in Brazil and at San Diego State University (US) as an exchange student. Currently, I am tax inspector of the State of Alagoas in Brazil working as a data scientist and data analyst. It is an achievement for me finding solutions for business problems using data products such as insights, dashboards, predictive models and business reports.

I work monitoring big companies that have tax breaks, verifying if they are satisfying their tax obligations in order to get these benefits, besides of detecting and building frauds signs in tax payments for all types of companies and situations. Likewise, I calculate constants that are used in determining tax of some products such as alcohol, wheat, gas using statistical methodologies. I also work in personal data science projects.

Skills

Languages programming and databases

  • Python for data science and webscrapping
  • R for statistical modelling
  • SQL

Statistics and Machine Learning

  • Machine learning (Classification, Clusterization, Regression, Learn to rank)
  • Statistical Inference (hypothesis tests and confidence intervals)
  • Sampling Survey
  • Time series
  • Survival Analysis
  • A/B tests
  • Learning to rank
  • Design of experiments

Data Visualization

  • Tableau
  • Power BI
  • Metabase

Other Skills

  • Git, Github, Cookiercutter, Virtual enviroment e Docker
  • Streamlit, Flask, Python API's
  • Cloud Heroku, AWS Amazon, Google Cloud Platform (GCP)
  • Linux

Experiences

2+ Data scientist and data analyst

Using statistics to create methodologies which determines the consumer selling price of a product to define its tax basis. Creating reports and dashboards to monitoring companies tax obligations and calculating how much they must pay in taxes.

Main results:
1) Create a metodology that almost double the tax basis of ethanol which increase tax collection of this product in at least 30%.

2) Create a metodology that define a fair tax basis for gasoline, diesel oil, kerosene and ethanol based on consumer selling price of gas stations.

3) Creating an Python program which calculate aggregate margin that were outdated of lots of products to increase their tax basis. These new margins increase tax collection in more than R$ 40 milions.

4) Creating a report and dashboard in python to calculate how much some specific companies must pay in taxes. These report resulted in more than R$ 30 milion in notifications.

2+ Data science projects

Building data solutions for business problems, close to real challenges faced by companies, using publicly available data from Data Science competitions, where I tackled the problem from the conception of the business challenge to the deployment of the trained algorithm in production, using Cloud Computing tools.

2+ Tax inspector

I currently work monitoring operational movements of big companies that have tax breaks, veryfing if they are satisfying their tax obligations in order to get these benefits. In case of those conditions are not satisfied, I calculate how much companies own and send notifications charging these values.

Business Analysis Manager (Dec 2020 - May 2021)

I lead a business intelligence team that work creating and managing data of Alagoas Treasury Department in order to spot tax irregularities and frauds.

Main results:
1) Lead the team to develop a dashboard to detect which companies paid less than what was declared in tax returns. This dashboard result in more than R$ 10 milions in tax notifications.

Data Science Projects

High Customers Identification

This is an end-to-end data science project which a clusterization algorithm was used to select most valuable customers in order to create a loyalty program called insiders. Also, a dashboard in Metabase was created to monitoring these customers and their metrics. This dashboard is hosted in a EC2 service provided by AWS.

Insurance Cross-sell

This is an end-to-end data science project which a classification algorithm was used to rank clients which would be interested in getting a car insurance. It was used machine learning Random Forest algorithm to sort these clients. The deployment was did in a AWS cloud server (EC2) and results was accessed through Google sheets.

Rossmann sales Prediction

This is an end-to-end data science project which predict sales of the next six weeks of Rossmann stores. It was used machine learning XGBoost algorithm to predict these sales. APIs and model was hosted in a EC2 server of AWS cloud. Predictions of each store can be accessed by users through Telegram as shown below.

Isketch website A/B test

This is an A/B testing project that was made to see if a new version of a sign up button in a website is better than current one. This A/B testing was made using a reinforcement learning algorithm called Multi-Bandit Armed (MAB) with a bayesian agent (Thompson Sampling).

House Rocket

This is an end-to-end project which helps a company finding good opportunities of buying houses and sell them later. It suggests what properties company should buy and if they need repair before sell them to maximize company's profit. A dashboard was created to show these properties and to advise the best opportunities.

Get in touch