Victor Augusto Silva Barros Serra

Responsible for studies and models in the logistics area of Americanas S.A., specifically regarding the definition and distribution of 12,000 items across 1,600 stores responsible for an average monthly sale of 600 million BRL.
Designed clustering models for stores based on income, geolocation, and store characteristics using SVM.
Developed sales prediction models for different departments using XGBoost combined with SARIMAX (Pytorch).
Created efficient product allocation models based on strategic company decisions in each area.
Designed the data flow in SQL (BigQuery).
Constructed and maintained reports in Looker Studio and Looker Plataform.
Built reports in SQL in Looker and in Python.
Created and automated models and studies of freight policy.
Generated campaigns.
Constructed managerial DREs.
Developed studies on pricing campaigns for critical stock.
Developed models to assess impacts on freight policy.
Conducted web scraping of e-commerce sites to simulate freight costs of products.
Developed LLM models to match a database of 80 million products with those captured from competing companies.
Created and maintained automated data models in SQL/Big Query.
Created and maintained dashboards in Power BI Premium for 400 employees across various departments.
Managed diverse reports using agile methodologies with 2-week sprints.
Proposed key performance indicators.
Conducted studies.
Generated numerous operational reports detailing information about 250,000 B2B clients within a Big Data environment.
Developed a logistic regression ML model to predict which items were likely to be delayed.
Developed a Naive Bayes ML model to forecast which business clients were at risk of leaving the platform.

Modulation Project (Current): Responsible for studies and models in the logistics area of Americanas S.A., specifically regarding the definition and distribution of 12,000 items across 1,600 stores responsible for an average monthly sale of 600 million BRL. I designed clustering models for stores based on income, geolocation, and store characteristics using SVM. I also developed sales prediction models for different departments using XGBoost combined with SARIMAX (Pytorch), created efficient product allocation models based on strategic company decisions in each area, and designed the data flow in SQL (BigQuery) and the construction/maintenance of reports in Looker Studio and Looker Plataform.
Freight Polices Project (2021/2022): Built reports in SQL in Looker and in Python, created and automated models and studies of freight policy, and generated campaigns. Responsible for constructing managerial DREs, studies on pricing campaigns for critical stock. Developed models to assess impacts on freight policy. Conducted web scraping of e-commerce sites to simulate freight costs of products. Developed LLM models to match a database of 80 million products with those captured from competing companies.
Marketplace – E-commerce project (2020/2021): Created and maintained automated data models in SQL/Big Query and dashboards in Power BI Premium for 400 employees across various departments. Utilizing agile methodologies with 2-week sprints, I managed diverse reports, proposed key performance indicators, and conducted studies. Numerous operational reports were generated, detailing information about 250,000 B2B clients within a Big Data environment. During this period, I developed a logistic regression ML model to predict which items were likely to be delayed, enabling the responsible team to
take proactive measures. Additionally, I employed a Naive Bayes ML model to forecast which business clients were at risk of leaving the platform.
Main tools: Google Cloud Platform - BigQuery (SQL Extraction); Confluence (Documentation); Jira (Scrum Methodology); JupyterNotebook (Python); Looker and Power BI Premium (BI).

Responsible for generating comprehensive reports using Power BI, QlikView, and SQL for the management of Vuon Card, a Private Label credit card.
Developed various studies and constructed indicators and goals to drive business strategies.
Conducted a detailed statistical analysis of customer lifecycle (cohorts).
Established minimum targets for new registrations needed to sustain sales over time.
Set benchmarks for growth.
Implemented k-means ML clustering to identify key groups of lost customers for targeted recovery efforts by the customer service teams.
Applied k-means clustering to categorize good debtors.
Facilitated strategies for increasing compensatory interest revenues.
Led to a significant revenue increase of 100,000 BRL from a baseline of 1.2 million BRL in average monthly revenue.

Developed a script in a Docker container to capture data from real estate and vehicle auctions in Brazil.
Sent data to Databricks.
Created reports in Power BI.
Scraped vehicle price data from official Brazilian databases.
Created inflationary indices.
Developed time series models using Temporal Fusion Transformers to predict future prices.
Utilized LLM (Large Language Models) for reading legal documents to parameterize automated calculations.
Used Python scripts to structure unstructured data.