Built core APIs around authentication, authorization, encryption key escrow, event ingestion, and external service integrations. Python, Django, Postgres, Redis. Retooled AWS infrastructure in both Terraform and Cloudformation. Implemented monitoring, metrics, APM, log aggregation, etc. Improved deployment turnarounds by building CI/CD pipelines. Jenkins, Ansible, Packer, ECR. Transitioned backend through product pivot, changing logic and infrastructure to handle larger orders of magnitude requests for analysis and aggregation. Refactored features in the Django monolith into scalable and maintainable Golang services. Built load testing tools that statistically measure the performance of the backend and infrastructure. Experimented with Kinesis, Spark, ElasticSearch, and Kubernetes as part of recent re-architecture efforts.

Built and maintained metrics data pipeline that batched, aggregated, and ingested 5 TB every day. Python, Golang, Postgres, Redshift. Optimized infrastructure to deal with bursty traffic upwards of 50k QPS. AWS, Terraform, SQS, Kinesis, S3. Engineered asynchronous schema migrator that allowed for consistent destructive and constructive schema changes across multiple ingestion services. Created internal dashboards to let data scientists trigger these schema changes in a safe and rapid way. AngularJS, HTML/CSS, Golang, Postgres. Implemented Modular API to query statistics from data warehouse for use by other services. Python, Flask, Spark, Golang. Updated internal data pipeline dashboards to better handle errors encountered in ingestion, sometimes automatically resolving them on a case by case basis.

Proposed, designed, and developed an Ad Campaign cost optimization ML model by analyzing user data to determine better CPC and CPA rates. This utilized globally scalable software, like Hive, Pig, and Spark, in order to filter and aggregate sparsely populated data fields in order to build a ML model.

Developed a MS-Access based user interface for a back-end Dataloader for Oracle E-Business Suite Projects Application. This interface accepted input from MS-Excel based data for various Oracle entities like Projects, Tasks, Agreements and Budgets. It enabled review of the data in a single/multi record blocks and on completion of the review, the data is pushed to Oracle interface tables. Dataloader processes are run to load data into the Oracle application. Successfully processed records are deleted from MS-Access and any error records are returned for correction and reloading.