Improving data platform robustness, scalability, reliability and internal user experience.Building compliant and secure unified data lakes.

The privilege to revisit my family, my roots.

Taking underwriting model prototypes to scalable production with fully supported A/B testing and monitoring infrastructures, observability and re-training pipelines.

Designing a fully customizable self serve machine learning deployment platform with shared data integrations for every business vertical within Block.

Developed Data ETL pipelines utilizing AWS Kinesis, Redshift and Spark.Established company-wide data access best practices and data warehouse workload management.- Proposed options for a performant data warehouse and fine-tuned BI tooling to reduce concurrent dashboard loads from over 1 minute to 12 seconds.- Deployed a prediction service through a Lambda function to estimate the USPS mail time-to-delivery. Improved the existing accuracy from 4% to 60%.- Created the company's first data pipeline by ingesting data from multiple sources such as Kinesis Streams, CSV uploads and PostgreSQL databases to create an aggregated view that computed partner-related SLAs. Reduced job execution time from 1 hour to 15 min.

- Designed and implemented the internal PDF rendering regression testing tool that inspected the visual discrepancies between PDFs. Integrated the tool into the standard deployment workflow to eliminate the need for human testing. Utilized native Golang concurrency to achieve average test latency of 100ms.- Designed and implemented a real-time logs collection system from a consumer deployed on partners' machines. Logs were designed to be persistent and to minimize negative effects on the throughput of the consumer.- While working on the API scaling initiative, created a containerized solution deployed with AWS to send a high volume of API requests for throughput testing.

- Designed and implemented the "Idempotent Post Requests" feature allowing customers to prevent thecreation of duplicate resource in an event of 50x server timeout errors.- Contributed to build a Machine Learning based solution for detecting fraud on the platform, reducing instances of fraud by over 99%.