Eloi – Python, Data Science, Data analysis

Customer had a deprecated web scraper for the major real state advertiser in the USA. They needed to design a better system, including refactoring the web scraper, defining and building a data architecture, and ETLs. There were also some analytical work and dashboard applications built on top of the extracted data.

Eloi was responsible for defining the roadmap, refactoring the web scraper, designing data architecture, and implementing the solutions on AWS.

A European fintech that offers a "buy now, pay later" service, among other key financial services.

It has two major cost factors in its operation:

1. defaulted orders (unpaid)
2. fraudulent orders

In order to help mitigate those, it tries to verify the identity of its new customers. Those verifications vary from market to market and, in some cases, yield low conversion rates (for legitimate customers).

• Identified an issue in the authentication process for new customers in the company's major European market (Germany), which was causing it to have the lowest conversion rate in Europe (30%);
• proposed and implemented NLP methods to fix the authentication process, successfully doubling (30→60%) the CR for legitimate customers.

This is the largest food delivery company in the western hemisphere, having over 60MM orders per month. At 2020, it had about 6000 employees. During the 2020 lockdowns, the company faced a new challenge - how to perform its yearly evaluation process for all its employees who were now working from a home office.

• Eloi and his team built an MVP of an ONA (Organizational Network Analysis) with data on employee interaction through Slack, Google Suite, and Google Meets that allowed them to pinpoint which employees should evaluate who. Furthermore, we integrated our solution into Slack - enabling a seamless and effective evaluation process across the entire company.
• Eloi also proposed a graph-based approach known as ONA (Organizational Network Analysis);
• built an MVP for the Data Science department (200 people);
• convinced stakeholders (CHRO) of the viability of the project;
• assisted in hiring and training a head of people analytics to lead the project;
• provided guidance and assistance on the development & deployment of a company-wide version of the project.

This is the largest food delivery app in the western hemisphere, having over 90% market share in Brazil and 60MM monthly orders. In late 2019, the company faced a fraud crisis, as overnight, its chargeback rate went from ~0.5% to over 3.5%. Operating at very thin margins, the company could go bankrupt if it didn't address this issue quickly.

• Eloi was moved from his original team (growth) to the anti-fraud department to work alongside other professionals as an internal consultant, diagnosing the problem, proposing a solution, and implementing a solution;
• he correctly identified the underlying causes of the spike in fraudulent activity;
• defined KPIs related to the different types of fraud;
• built monitoring & reports on those KPIs;
• refactored the anti-fraud decision engine rules (a system with a set of rules that evaluated the legitimacy of each transaction in real time);
• defined and build new protocols, processes, and systems to mitigate fraud further;
• created and automated a friendly-fraud dispute process;
• hired analysts and data scientists;
• implemented an agile methodology;
• managed stakeholders' expectations and department roadmap.

It was an incubated startup at Google Campus - São Paulo. It aimed to enter the heavy freight logistics market in Brazil (which is mostly composed of truck deliveries). As a starting point, it was using publicly available data in its manual operation. The company needed to be intended to collect that data from established competitors and leverage data and AI to disrupt the market.

• Made a web scrapper for publicly available data;
• created data & system architectures on AWS;
• built a GUI for the operations team to access and utilize that data.

An EduTech that intended to bring technology and data intelligence to the public education system in Brazil. Its app offers different features and services for parents, kids, and teachers in state schools.

One challenge it faced was its new feature, an in-app answer sheet for teachers to use during exams, which allowed for autocorrection and grading. The issue with this feature is that a small percentage of students didn't have smartphones, and those still needed to have their results uploaded and integrated into the system. If there were no solution to it, teachers would prefer to do the whole exam manually than do it partially manually, which would yield a drop in the adoption of the solution.

• Created an OMR (Optical mark recognition) program that allowed for the user to take a picture of a paper answer sheet and have it automatically graded and assigned to the correct student.