Saurabh Agrawal

I lead the Financial Engineering organization at Coinbase, where we’re building the high-fidelity systems that serve as the bedrock of trust for the crypto economy. Since joining, I’ve built a world-class global team across the US and India from the ground up, redefining our technical North Star to support hyper-growth. I’m a hands-on leader who believes in fixing the foundation while building the skyscraper, ensuring our infrastructure stays miles ahead of our aggressive global expansion. My scope covers the most high-stakes pillars of financial integrity: I own the global subledgering, payment reconciliation, and safeguarding platforms that act as the single source of truth for billions in transactions. Beyond the ledger, I drive the strategy for our Tax Platform, which handles over 2 billion line items via a modular data factory, and our AI-native Tooling Platform, where we’re replacing fragmented legacy tools with intelligent, multi-agent AI systems. We don’t just move data; we build the immutable, high-throughput systems that define financial precision on a global scale.

Amazon Last Mile Technology team is responsible for envisioning and developing cutting edge initiatives in delivering packages to our customers as fast and as cost effectively as possible. We build, own and manage hundreds of web scale software services and systems, that enables package delivery for billions of parcels throughout the world. We are hiring for SDMs, SDE2, SDE1 positions in Gurgaon & Bangalore.

Adobe FrameMaker * Planned, architected and executed the delivery of FM-64 bit migration including in-depth knowledge building of FM internals for understanding the scope of each change * Complete ownership of the agile development cycle including sprint planning, managing JIRA dashboard, SCRUM meetings * Developed in depth understanding of the compilation process including code generation, static linking, dynamic linking, makefile, Visual Studio compiler, dumpbin etc. * PDF Library: Replaced the layout and PDF writing process to integrate a new and better PDF generation library FrameMaker Publishing Server * Adobe FrameMaker Publishing Server (FMPS) is an enterprise software that allows you to automate your multichannel publishing process. * Re-architecting the complete solution to make it OS agnostic (run on Linux, Windows etc.) * The solution now scales horizontally to achieve unlimited throughput

DITA XML Add-On (CCMS) for AEM, Adobe Systems * Adobe Experience Manager is a comprehensive content management system solution. Its plugin architecture allows creation of customized solutions. DITA XML Add-On was created to support managing, authoring, reviewing and publishing XML files conforming the DITA standard. * Product was built from start by a team of two people. Initially, it was supposed to be a POC only. However, the response and delivery was very smooth and it was converted to complete product in less than six months. * AEM Connector: Created a CMS connector in FrameMaker using Java/C++ which would allow user to view the directory structure, upload files, edit existing files and manage existing versions * AEM Workflows: Created workflows in AEM using Java which allows user to initiate reviews, suggest authoring changes and publish XML documents to multiple formats like PDF, HTML5, AEM sites, EPub etc. The web interface was developed using JavaScript/HTML/CSS

* Simplified Xml Author is an authoring environment within Adobe FrameMaker which allows user to create XML files without worrying about the actual XML grammar. * Templatized Authoring: Template designers could create a template document with headings and banner text which would help a subject matter expert write his views/findings * Guided Authoring: Guide a user while authoring, using non-intrusive messages, to help complete a document * Valid XML creation: During authoring, valid XML was created despite the user not knowing about the XML grammar and structure. Hence, he could focus on his area of expertise and document becomes ready for other workflows like publishing, sharing etc.

* FMPS is an enterprise software for publishing multi-channel outputs from single source. * Task manager and scheduling service: Evaluated and developed understanding on WCF Framework (Windows Communication Foundation) and Java Spring Web MVC. Created a web service using the WCF FrameWork following the REST protocol which allowed user to manage tasks and publishing schedules per the CRUD operations in C# .Net. Handled inter-process communication and various race scenarios * Task executor: Created a tool in C/C++ which enabled execution of tasks created using Web Service. Executor had to interact with Adobe FrameMaker APIs for achieving various tasks like connecting to CMSs, publishing, uploading files to server etc. * Frontend: Created a web interface which interacted with the Web Service to create/manage tasks using JSON/XML API response. Developed using Google Closure Framework in JavaScript, HTML, CSS

* As an individual contributor, developed and worked on large set of features for two major releases of FrameMaker (10.0 and 11.0). * Document catalogs: Built and improved catalogs (Character catalog, paragraph catalog and table catalog) for ease of authoring * DITA Authoring: DITA is a popular XML standard for technical documentation. Added many features and workflows to support this new standard with FrameMaker * Optimize output generation: Researched extensively with performance analysis tools like Visual Studio profiler, Intel VTune etc. and optimized PDF generation process by more than 80% * Banner Text: Banner text in a document instructs user about what to enter in the document. This text, which is customizable, would disappear once the user starts typing. Worked extensively on text layout, line breaks etc.

Information Extraction: Attribute extraction with Conditional Random Fields using domain specific knowledge. Deduction of product attributes like Title, Price, Image and Description from shopping web pages

BTech and MTech in Computer Science from IIT Delhi IIT Delhi Thesis: Indoor Robot Navigation with Single Camera Vision * Automatic navigation of robot while avoiding obstacles to reach a specified destination * Map generation and learning to obtain a relational graph; and path planning to find the shortest path