Jul 2022 – May 2025 · 2 yrs 10 mos · Silicon Valley, CA

• Designed and implemented SSL certificate management to optimize SSL termination outside the Kubernetes cluster, overcoming inefficiencies of the previous ACME-based approach for Meta’s Conversions API Gateway, a product used for processing user events for effective advertisements. Implemented a solution to handle SSL termination outside the cluster while supporting 5,000 domains. The load balancer would support a domain hosted in Cloudflare which in turn handled additional domains as custom hostnames on Cloudflare's fallback origin server. This enhancement improved SSL handshake efficiency and freed up cluster resources for core application logic. The project resulted in 10x increase in events processing capacity compared to the previous architecture, resulting in 10x reduction in costs for our ad agencies, making our platform more cost-effective and accessible specially to small business advertisers.
• Designed and developed an autoscaling system for Meta’s Conversions API Gateway. The system allowed our cluster to dynamically adjust resources, efficiently handling surges in events and ensuring optimal performance during peak demand
• Streamlined and implemented feature flags, facilitating targeted feature enablement for specific clients before widespread deployment to enhance controlled feature rollouts.

Oct 2020 – Jul 2022 · 1 yr 9 mos · San Francisco Bay Area

• Designed and developed a multi-threaded system which receives messages in the range of ~100M daily from SQS messaging queue, parses each message, stores it as a document in Elasticsearch cluster. Developed a separate system which reads the documents and creates a report out of it. Such a report enables customers to check the status of each product update
• Hackathon project: Designed and developed a system that receives stream of real time product updates in ION format, parses it, streams to Kinesis Data streams in JSON format which is consumed by Kinesis firehose using Glue for schema reference and stores in S3 in Parquet format for easy and efficient query from customers
• Independent project: Designed and developed a system to copy DDB data to parquet files in S3 using Glue and created Athena table to query this data using SQL. This provided easier way for product managers to understand customer usage of the service
• Software Skills used
• Programming Language: Java
• Frameworks: Dagger, Guice, Junit, TestNG
• Storage/Databases: DynamoDB, S3, Elasticsearch, ElastiCache(Redis), MySQL
• Other: Microservices, Spark, Mapreduce
• Cloud: AWS – EC2, Lambda, API Gateway, SNS, SQS, Step Functions, Kinesis, Glue, Athena, IAM, EMR, Cloudformation, Cloudwatch

Apr 2017 – Sep 2020 · 3 yrs 5 mos · San Francisco Bay Area

• Product information analysis: Designed and developed a system which processes information of products (like number of products at any given time) we see on Amazon's website, arranges this infomation in a particular format and publishes this information to an S3 bucket with metadata so that other teams can use this data for graphical analysis (like knowing whether counts of products is increasing in reliable way or not).
• Designed and developed a system which deletes S3 objects at specific time resulting in savings of $25Millions in 5 years

Apr 2016 – Mar 2017 · 11 mos · Greater Boston Area

• Developed Pixel Management System for advertisement of recently seen items on Wayfair's website by a customer.
• Analyzed and processed real-time streaming data using Storm after reading the data from Kafka topics and then sent to the vendors. This data is generated from activities on website and mobile app
• Built the functionality for browsing rugs using visual search backend. User could enter the URL of the image and similar rugs result is shown to him and implemented filter functionality for rugs browsing.
• Recognized as ‘Solution Superstar’ (Aug 2016, Dec 2016) and ‘Internal Innovator’ (Dec 2016).
• Technology Stack: Java, PHP, MS SQL Server database, JQuery, Apache Storm.