Data Science at Tapsi

Between October 2020 and April 2022, I worked as a Data Scientist at Tapsi, one of the leading ride-hailing platforms. Working in the fast-paced mobility sector meant dealing with massive, continuous streams of data—the platform processed approximately 500,000 rides per day.

My work centered on building robust machine learning pipelines to protect the platform’s integrity and optimize its growth strategies.

Fraud Detection

In ride-hailing, malicious actors constantly find new ways to exploit the system. I developed and maintained machine learning pipelines dedicated to the automatic detection of fraudulent drivers.

To achieve this, I profiled both driver and passenger characteristics based on their ride history and referral information. I mapped these profiles into a complex network graph, analyzing the connections and behavioral patterns between users. By leveraging these graph-based profiles, the model could accurately estimate whether a given ride was fraudulent.

This approach was highly effective at identifying sophisticated scams, particularly:

• Commission Evasion: Detecting coordinated efforts between drivers and passengers to bypass platform fees.
• GPS Spoofing: Identifying fake rides generated through manipulated location data.

Passenger Campaign Optimization

In addition to fraud detection, I contributed to the platform’s growth and marketing efficiency, specifically to optimize passenger marketing campaigns. By analyzing user behavior, ride frequency, and churn probability, we targeted campaigns more intelligently, successfully maximizing Return on Investment (ROI) as our primary key performance indicator (KPI).