Work
Selected work, presented as anonymized case studies. Employer and internal details are withheld; the problem, the design, and the outcome are real.
Rule-driven debt-collection engine
Worked on the orchestration core for a multi-channel debt-collection engine on a Southeast-Asian fintech lending platform. A rule-driven pipeline decides which collection action each delinquent loan should receive, on which day, across desk-call, field-visit, and messaging channels.
GoTemporalKafkaPostgreSQLCQRSRedpanda Connect
Taming memory in a high-volume event pipeline
Stopped a stream-processing consumer from running out of memory on the heaviest events, by stripping an unused multi-megabyte field at the byte level before any JSON parser allocated it. The same idea applies to every Go consumer in the fleet.
GoRedpanda ConnectBloblangKafkaIBM/sarama