Scaling Bulk Token Enrollment
Resolving stuck card-token uploads and redesigning the workflow to process 10,000 records in under one minute.
Customer context
Banks and card-issuing partners used bulk file uploads to enrol large volumes of card tokens on the SafeToken authentication platform.
The problem
Under heavier workloads, some uploaded files became stuck during processing. This delayed token enrolment, made completion difficult to predict, and limited the progress information available to product owners and customers.
My role
I investigated the processing bottleneck, redesigned the bulk-upload architecture, aligned the proposed solution with technical teams and product owners, and implemented the improvements.
Product owners managed customer communication while I provided technical findings, implementation updates, risks, and progress reports throughout the work.
Technical redesign
- Divided large uploaded files into smaller processing batches
- Published batches through Kafka to decouple file intake from execution
- Introduced controlled parallelism so independent batches could run concurrently
- Isolated failed work so a single batch could not block an entire file
- Improved progress tracking, retry behavior, and operational visibility
Outcome
- Resolved the stuck-file processing problem
- Processed 10,000 token records per upload in under one minute
- Created a more scalable and resilient enrolment workflow
- Provided clearer progress information for technical teams, product owners, and customers
What this demonstrates
The work combined customer-impact investigation, distributed-system design, stakeholder alignment, implementation ownership, and progress management. It solved the immediate customer problem while addressing the architectural limitation that caused it.