Antonio Petrov

Led a team to develop a comprehensive data archival system for Societe Generale Expressbank (SGE) following its merger with another bank. The system serves as a unified platform for viewing an array of transactional data, consolidating information from multiple databases into a single Oracle database.

Developed a comprehensive subscription and device management platform with a focus on responsive design and mobile-first user experience. I collaborated closely with a UI designer on design assets and another developer on back- and front-end development.

A project for a medical company for generating documents (payment forms) and management approvals. Due to the importance of the documents, each one would have to be personally approved by multiple people following a specific workflow. The workflow is managed by the administrator without requiring a developer to jump in and edit anything.

To create a captivating and intuitive landing page that gauges the level of interest in a forthcoming SaaS tool designed for tracking commissions and gamifying sales processes. The aim was to validate the market demand for such a solution before full-scale development.

Optimized a React/TypeScript app by migrating from Firebase to Supabase. I transferred user authentication, database, file storage, and edge functions to Supabase for better performance. I also implemented pagination to resolve dashboard loading timeouts caused by excessive data loading and enhanced the app with complex filtering capabilities requested by the client. I fixed minor bugs and updated Survey Monkey webhooks to integrate with the new Supabase edge functions. The migration and optimizations significantly improved the app's efficiency and user experience.

For one of the web applications I was working on, I had to develop an integration with SQS (FIFO) to process messages as quickly as possible. To achieve a very high processing speed, I used multithreading across multiple batches. The number of batches is editable through the database. This allows the admin to easily regulate batch processing amounts without needing to redeploy every time a setting changes.

I worked on a web application for internal use where staff members would upload scanned files in PDF format, and we would use OCR to process them by parsing the data, putting it in a database, and in different S3 buckets based on certain filters and keywords that the files would contain. The staff would upload these files all day long, but the rate at which they would upload these files would constantly change, from a few files per hour to hundreds of files per minute. To make sure that the processing speed remained consistent, I used Kubernetes to allow for automatic scaling as needed.

Documents were pushed into RabbitMQ, and worker pods consumed them. We used KEDA with the RabbitMQ scaler to automatically scale the worker deployment based on queue length. Since OCR was CPU-intensive, scaling the worker replicas allowed us to handle traffic spikes efficiently and keep the queue lag under control. The upload process would make use of CQRS events, where a DocumentUploaded event would be submitted when a user uploads a document.