I was deeply involved in architecting and implementing the big data workflow using Spark and Scala, which allowed massive parallel processing to meet the seven-to-eight-minute window. Reconciling GPS coordinates from satellites with data from OBUs sometimes required advanced interpolation techniques due to inaccuracies or missing signals.

Additionally, robust logic had to be implemented to check for anomalies from toll gates, ensuring that single sensor glitch did not invalidated toll charges.

An important facet of my role was handling analytics and metadata: although the raw data had to be wiped clean on schedule, we needed derived metrics for auditing and fraud detection.

By preserving selected aggregated statistics, I implemented multiple algorithms which could identify suspicious patterns, such as repeated OBU “offline” episodes or contradictory location evidence.

These flags were instrumental in uncovering attempts to manipulate the toll system, helping enforce proper road usage fees. In the process, we effectively replaced thousands of scattered Excel macros, standardizing calculations in a single, reliable application.