Serverless Functions (Async Tier) – Meta

Async Tier is Meta’s distributed job orchestration platform, executing over 50 trillion tasks daily across 20,000+ job types and supporting 4,600+ teams on 100,000+ servers. It enables petabyte-scale data processing and powers mission-critical workloads for products used by billions—including Facebook, Instagram, WhatsApp, Ads, and AI initiatives.

• Privacy Compliance Leadership: Drive and achieve Privacy Infra Certification. This milestone reduced onboarding friction, streamlined compliance, and established Async as a privacy-first platform.

• Reliability & Efficiency Improvements: Drove systematic reductions in OutOfMemory (OOM) and Timeout errors, lowering unhealthy job types from 16% to 12.5%, reducing OOM failures by up to 70%, and delivering $400K–$1M in annual compute savings.

• Advanced Resource Management(Traffic Shaping): Launched adaptive constraint based traffic shaping feature in async, onboarding 80+ async jobs, freeing up $2.86M/year in GPU resources, and improving model coverage for key AI and data teams.

AWS SageMaker(a fully managed AI/ML platform) — MLOps Infrastructure

Worked on core infrastructure components of AWS SageMaker(a fully managed AI/ML platform) Pipelines, enabling customers to build, orchestrate, and automate end-to-end machine learning workflows in production environments.

• Led the design and implementation of a critical enhancement (Selective Execution in Sagemaker Pipeline - https://docs.aws.amazon.com/sagemaker/latest/dg/pipelines-selective-ex.html) and optimizing the MLOps lifecycle for AWS SageMaker pipeline resulting in 750 days of ML builder effort and 7000 hours of execution time saved annually.

Amazon Finance Data Lake

• Played a foundational role as a core engineer in establishing Amazon’s expansive authoriatative financial data lake initiative, contributing from its inception and leading its development from the ground up.

• Orchestrated the expansion of dataset ingestion services, increasing the number of datasets managed from 40 to 1200 and reducing ingestion time from 2 hours to 15 minutes.

• Implemented self-service dataset onboarding, cutting onboarding time from 5 weeks to 2 days.

• Scaled data lake from 100TB to 8PB, enabling data-driven decisions for 300+ teams.

• Achieved $6M+ annual savings by optimizing infrastructure.

I was awarded Seaborg Summer Research Fellowship, 2017 by the Seaborg institute, LANL.

• FESD - (F-Electronic Structure Database)

– Designed a fully automated Density Functional Theory (DFT) simulation based data generation system using python scripts for f-electron compounds.

– Streamlined data processing techniques by utilizing the HPC clusters more effectively, that improved execution time for the data processing step by approximately 4x times.

– Developed a JSP-Servlet based web app with interactive visualization tools to visualize these scientific data.

– Built several query tools to explore correlations and trends in f-electron compounds.

– Technologies: Java, JSP, Servlet, Mysql, Bootstrap, Javascript, Ajax, Python, Selenium, PBS, HPC, Plotlyjs

Worked on projects as a Full Stack software developer for clients like Dell, Bestbuy using Agile methodologies.

• Brandshare | Marketing Operation Management System

– Developed Asp.net, AngularJS and SQL Server based web application.

– Worked on developing several features like activity workflow management, automate content authoring, content authorization, report generation, automate data warehousing tools etc.

– Integrated Amazon CloudFront, Amazon S3 etc to accelerate the content delivery of the Brandshare sites.

– Worked on numerous migration tools, prototypes, proof-of-concepts and integrated several 3rd party tools.

– Technologies: C#, Asp.Net, MS SQL, CSS, HTML, JavaScript, JQuery, Asp.Net web api, AngularJS 1.0,

knockout js, Bootstrap, Kendo UI, Telerik Platform, IIS, Git, AWS Cloud services like S3 and CloudFront.

• Brandshare Automated Deployment Tool

– Built a comprehensive deployment tool with Go-Continuous Delivery System to organize, build and execute deployment packages.

– It reduced 50% manual interactions in the deployment process and ensured a fast and error free deployment.

– Technologies: C#.Net, MongoDB (No Sql), Go-Continuous Delivery System, Nant build tool.