I work on Amazon Catalog Metadata Systems. Catalog Metadata describes all the product requirements, validation, and processing rules for catalog content on Amazon. The catalog metadata systems have 6 different implementations that have been around since 2012. Much of my focus has been on unifying the 6 different implementations in order to speed up developer velocity. I also re wrote the processing engines memory model on three of them in order to launch a new program. The program was to expand amazon's product types from 100 to 10,000 product types. The expansion of product types helped grow selection and quality for the catalog. The cardinality explosion of metadata wouldn't scale on our legacy systems and hence my rewrite. My rewrite of the engine had a nice side effect of dropping latency from 500 ms to 100ms for sellers managing their listings on Seller Central website.

I worked on logistic software for sorting packages within Amazon's logistic network. This included Fulfillment Centers, Sortation Centers, Delivery Stations, and Air Gateways. 40 Tier 1 micro services were transferred from a dev center in India to Tempe, AZ in 1 month due to high attrition in India. I lead the knowledge transition and stabilized operations with new owners in Tempe. While transitioning and building new teams, we also delivered key features for Air Gateway launch to weigh and load balance Amazon Prime Air planes. I also helped launch human safety software for amazon truck drivers and associates to warn them when there are dangerous trailers. This software helps safe human lifes.

I worked on building micro service blueprints for application teams in Amazon Seller Central. I prototyped and built a few blueprints with AWS CDK (Typescript). Blueprints encompasse, CI/CD, Compute, VPC, Monitoring/Alarming. The blueprints I built had three compute choices (lambda, ECS/EC2, ECS/Fargate). The CDK blueprint allowed new microservices to be provisioned with 10 lines of typescript. The blueprint was used for 30+ microservices across 5 teams in the org.

Product Development team for AWS Cloud Watch Logs Insights. I lead a group of engineers working on the data plane for ingesting logs from AWS customer and other AWS products, some examples include AWS Lambda logs, VPC Flow logs, and customer service logs the internal format of Amazon. My contributions to the project were mostly around changing the ingestion architecture in order to scale to ingest all existing logs traffic and to handle high voiume quick accelerating AWS Lambdas. Our IAD region only at launch ingested, 100PB of log data / month. The product's auto scaler had to provision log ingestion for customer's Lambdas that would accerlate from 0GB to 2GB of log data in 3 minutes. The product launched internally 4 weeks before AWS RE:Invent and then publicy at AWS RE:invent.https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/AnalyzingLogData.html

I worked on Seller Central's backend systems for processing price, inventory, product, image, and shipping updates from third party sellers. The system processes 100K TPS for amazon catalogs. In my first year, I moved a few of the systems Oracle relational data stores to Dynamo DB. In my 2nd through fourth year, I designed, implemented, and test a platform to move all of Seller Central's Feed and Interactive updates to a Kinesis + S3 + In memory system. The product launched and was product group moved to London in 2018. It is currently the orchestration system/platform for all 3P Sellers and Vendors who are updating the Amazon Catalog.

I worked on a Quality of Service platform which provides real-time telemetry data to Windows and Windows Live operation and product groups including OneDrive, Outlook.com, and Skype. The platform consists of a data collection runtime, data processing, and data warehousing. The product helps operation groups reduce their time to detection/resolution for service outages. I helped lead an infrastructure feature crew on this team that worked on scale and infrastructure problems. As a leader on the feature crew, I mentored junior level developers and interns.• Worked on “Lights Out” service management for our product at scale. This included automatic deployment, monitoring, and repair. Our solution managed around 1000 servers in 10 datacenters which include stateless, semi-stateless, and state full servers.• Developed/Designed a scalable, fault tolerant architecture for running background database maintenance jobs. These log running jobs would run reoccurring jobs for data expiration, data aggregation, and indexing. • Improved/Maintained our data processing severs and our databases. I have investigated/debugged performance issues with data processing. I have made data driven decisions on picking areas of our processing pipeline to optimize.• Automated partner provisioning which required multiple databases to be built and configured. The manual process took a business week while automation took under 20 minutes.

Worked in a lab that built sea mine detection systems for the NAVY that included the development of real time embedded software, specifically querying for software versions and an LED light up sequence on start up. Also, worked with test software for these systems that included collecting performance counters and writing a feature to save a camera’s output to a MPEG file. I was exposed to working with C and C++ while working with Arete.

Worked on NIKE .NET, a web application, which sells Nike products to distributers. Specifically worked on Java based web service for managing clothing catalogs. Implemented a server control to sort and search products.