Client Background

  • Client: A leading travel firm in Singapore
  • Industry Type: Travel and Transportation
  • Products & Services: Cab services
  • Organization Size: 2000+

Our Solution

We developed an automated, serverless solution to parse toll statement PDFs from multiple sources (Linkt and NSW Government), extract relevant trip and billing data, and cross-reference with Drive Mate trip records to identify applicable trips. The solution dynamically computes trip mappings and updates them as new listings are added.

Solution Architecture

Deliverables

  • PDF parsing service deployed using AWS Lambda
  • Integrated pipeline to match toll entries with trip and listing data
  • Storage of parsed and enriched data into a centralized database table
  • Trigger-based re-evaluation mechanism for unmatched entries
  • Dockerized script deployed through AWS ECR

Tools Used

  • AWS Lambda
  • AWS S3
  • AWS ECR
  • Docker
  • PostgreSQL
  • PDFPlumber / PyMuPDF (for PDF parsing)
  • MySQL

Language/Techniques Used

  • Python (PDF parsing, logic handling, database interaction)
  • SQL (trip and listing joins)
  • Docker for environment packaging
  • Event-driven programming with S3 Triggers

Model Used

Not ML-based; logic-driven model for data parsing and time-based trip matching using structured joins.

Skills Used

  • Python scripting
  • Dockerization
  • Event-driven cloud architecture
  • ETL pipeline development
  • Data modeling and SQL optimization

Databases Used

  • PostgreSQL (sync_db_au_prod, table: host_toll_statements)

Web Cloud Servers Used

  • AWS Lambda
  • AWS S3
  • AWS ECR
  • AWS CloudWatch (for logs and monitoring)

Technical Challenges Faced During Project Execution

  • Handling different formats of toll statements (Linkt vs NSW Gov)
  • Parsing unstructured text from PDFs reliably
  • Matching toll data to Drive Mate trips without a direct unique identifier
  • Ensuring idempotency and atomicity in trip matching logic
  • Delayed matching due to missing tag_number in initial uploads

How the Technical Challenges Were Solved

  • Used custom parsers tailored for each company’s statement format
  • Implemented robust regex and text extraction techniques
  • Cross-referenced tag numbers with publicData.deviceManagement.tolltag.tagNumber and matched with trip data using timestamp logic
  • Created fallback triggers to reprocess unmatched entries when new listing data is added
  • Deployed through Docker and Lambda for auto-scaling and event-driven execution

Business Impact

  • 100% automation of toll parsing and trip identification
  • Reduced manual intervention and data entry time by over 90%
  • Enabled real-time integration of toll data into Drive Lah’s backend systems
  • Improved trip billing accuracy and transparency for users
  • Streamlined operations and enhanced the customer experience

RELATED ARTICLESMORE FROM AUTHOR