Platform Architecture, Security & Reliability

• Separated PII tokenization and job management from the main backend into dedicated NestJS/TypeScript microservices, improving separation of concerns and limiting unnecessary access to sensitive data.
• Evaluated internal and third-party SSN tokenization options across infrastructure, engineering effort, functionality, and cost, recommending a provider that met security and functional requirements. Built the service using TDD and deterministic tokenization to detect duplicate SSNs, restricted de-tokenization to authorized staff, and recorded activity in WORM-protected S3 audit logs. Delivered the service to staging within one week.
• Designed auditability and result-generation controls for the weekly promotional rewards process so results could be reproduced and verified as unaltered during regulatory or audit review. Implemented tamper-evident records in WORM-protected Amazon S3 and independently verifiable randomization using AWS KMS and external entropy sources including NIST and DRand.
• Reworked scheduled-job processing with an AWS Step Functions/ECS workflow, adding standardized success/failure reporting and exponential-backoff retries. Applied the framework to the roughly eight-job weekly scratch-off process so processing could stop on critical failures, alert the on-call engineer, and resume from the failed step after correction, significantly reducing incomplete jobs.

Integrations, Delivery & AI

• Implemented automated reconciliation and monitoring for missing and anomalous data between Opencash and its clearing firm using webhooks and scheduled SFTP feeds. Added trade-order error handling and reliability improvements to keep internal records aligned with clearing-firm data.
• Automated TypeORM database migrations during staging and production deployments using AWS CodeBuild as a GitHub Actions runner inside the VPC, removing a manual step that could be missed, run against the wrong environment, or leave the application and database out of sync.
• Built an OpenAI API-based support chatbot for application questions and general trading terminology. Compared AI options and used Structured Outputs, Guardrails, and prompt instructions to keep the chatbot from providing financial advice, reducing support calls and emails.
• Simplified React Native/Expo releases with a GitHub Action that generated a pull request showing all changes since the prior release. After approval and merge, the workflow automatically started the Expo release, reducing release errors and making the process easier for newer engineers to learn and run.