Adhitha Dias

Adhitha Dias

Graduate student at Purdue University

Experience

Graduate Research Assistant, Purdue University, US

(Jan 2021 - Present)

  • SparseTDA: Accelerating Sparse Tensor Computations Using Compute Dis-Association With Fusion (June 2022 - Present)
    • Compile time optimizations for complex sparse tensor algebra multiplications using tensor kernel dis-association/distribution and fusion (ongoing).
  • SparseLNR: Accelerating Sparse Tensor Computations Using Loop Nest Restructuring (May 2021 - May 2022)
    • Implemented a new algorithm to change the loop nesting structure in code generation for sparse tensor computations.
    • The approach reduces the time complexity of the computation while maintaining a low memory footprint to better utilize the cache.
    • The introduced novel algorithm can obtain can achieve better speedups by orders of magnitude.
    • The paper written won the best paper award at the International Conference on Supercomputing 2022.
    • Tech stack: Tensor Algebra Compiler (TACO), C++, CMake, PAPI Cache Stats, Latex

Software Engineer, Sysco LABS, Sri Lanka (Branch of Sysco Corporation, Houston, TX)

(Feb 2019 - Dec 2021)

  • Merchandising Authorization and Authentication (Jun 2020 - Dec 2020)
    • Security concerns related to merchant portal access are handled by the integration of layers of security.
    • Carried out a majority of the work on the API implementation.
    • Technologies and Integrations: Microsoft Active Directory, Amazon Cognito, JWT, SSO, Introspection, Redis, PostgreSQL, Java, Spring Boot, Python, Liquibase, Spring Data JPA.
  • Merchandising User Management (Jun 2020 - Dec 2020)
    • Merchandising User Management application enables administrators to manage user access to the merchant portal.
    • Implemented the first cut version of the application.
    • Designed front-end, back-end, database, the entire stack.
    • Technologies: Spring Boot, JavaScript, Java, NodeJS, PostgreSQL, CSS, HTML
  • Enterprise Menu (Jan 2020 - Jun 2020)
    • Enterprise Menu is a Graph-based menu management system for multi-location restaurants. This enables different branches of a restaurant chain to maintain a variation of a single master menu with categories, items, modifiers, and taxes.
    • Work on this component included the use of technologies in the front-end, back-end, database, and deployment. Implemented the most complex sections in the enterprise menu component including but not limited to items, modifier groups and modifier options and integrations to POS/OLO applications.
    • Technologies: JanusGraph, Neo4J, Amazon Neptune Graph DB, NodeJS, ExpressJS, ES6, React/Redux, HTML/CSS, Gulp, Webpack, Mocha/Chai, Java Spring MVC, Wiremock, REST, Jira, GIT, Microservices, SonarQube, Jenkins, GrayLog, AWS SNS, AWS S3.
  • Communicator API (May 2019 - Nov 2020)
    • Communicator API facilitates email, SMS and voice message sending functionalities and is used by many other components in the organization.
    • The component features advanced queuing and failure notification alert mechanisms, which also consist of a dashboard view for internal use and testing.
    • Introduced a lock mechanism to distribute the workload among a cluster of instances for sending email SMS and voice messages and various other improvements to the application like dockerizing the component and deploying to ECS.
    • Technologies: MySQL, Twilio integration, SendGrid integration, Slack integration, Bash shell scripting, Redis, Graylog, Symfony, PHP
  • Sales Demo Data Pumping Tool (Cloud Reports) (Jan 2019 - Jan 2020)
    • Developed the application from scratch by myself.
    • The application consisted of a UI automation and a data pumping tool to mock an existing user and pump data as a different user by changing data that needs changing. This application was helpful for sales representatives for demo purposes of the Cloud Reports components.
    • Technologies: JavaScript, NodeJS, Puppeteer, Java, CouchDB
  • UA-REPORTS (Cloud Reports)
    • Engaged in feature developments in the application mainly in transaction reports section.
    • Universal Admin Reports is a modern reporting application used to provide comprehensive reports for restaurants using CAKE POS devices. Users can view data on sales, payments, transactions, payroll, and taxes and use advanced filtering options to analyze data.
    • The specialty of the application originates from its ability to handle enormous amounts of data for user accessibility.
    • Technologies: NodeJS, ExpressJS, ES6, React/Redux, HTML/CSS, Gulp, Webpack, Mocha/Chai, SonarQube, Jenkins, Graylog, Redis, REST, Docker, AWS ECS, AWS S3, Java Spring Web Flux, Wiremock, REST, Jira, GIT, Microservices, SonarQube, Jenkins, Agile.

Research Assistant, School of Information Systems (SIS), Singapore Management University (SMU), Singapore

(June 2017 - Dec 2017)

  • Wi-Fi based Indoor Localization using Distributed Antennas
    • Worked on finding a localization solution using different Wi-Fi access points using the angle of arrival of the Wi-Fi packets.
    • Technologies/Hardware/Algorithms: FPGA, GNU Radio, IEEE 802.11 PHY, OFDM, WARP devices, MATLAB, Music and SAGE algorithms
  • Follow My Lead
    • Video-based indoor navigation system.
    • Came up with a new algorithm for check-point acquisition from the video streams recorded.
    • Technologies: Android, AR, Sensor Fusion, OpenCV-based Image Processing