Research Assistant @ North Carolina State University
January 2022 - Present
Expertiza - Computer-Assisted Peer Assessment
Researching how peer assessments affect learning by employing Expertiza - an open-source peer
learning platform.
-
Improved peer reviewing process by transitioning from Generalized to Role-Based Reviewing
that provides skill specific evaluation metrics for assessment
of collaborators as per their roles.
-
Lead contributor in migrating legacy code from Rails 2.3 to Rails 6, modernizing a 10y+
old system without affecting any functionality.
-
Increased total test code coverage by 30% by writing automated test cases using RSpec
and Cucumber.
-
Identified and optimized process hindering logic by implementing optimal design
patterns, resulting in 20% performance improvement.
Full Stack Developer @ IBM
June 2019 - July 2021
Albertsons Companies is leading food and drug retailer that operates stores across 34 states
with more than 20 well-known banners.
- Developed and deployed “Schedule & Save” feature in production that lets shoppers set up
automatic orders.
- Built reusable front-end components on Angular 8 and backend microservices on Spring
Reactive for asynchronous order transactions.
- Ensured 90%+ coverage and fault tolerance by developing custom component and integration
testing framework built on JUnit5, Mockito, MockWebServer and SonarQube.
- Developed internal CI/CD Jenkins pipeline that integrated code build, unit testing,
sonar scan, and application deployment over Azure. Virtual Machines in a single scan
thereby reducing deployment time by 70%.
Westpac - Cards and Payments
Westpac is Australia’s oldest banking institution and financial services provider with more
than 14 million customers.
- Automated manual data collection and aggregation by dash boarding customer's verified
income and expenses data using Spring Boot microservices and HBASE clusters to aid
upfront customer background verification, resulting in
30% FTE reduction.
- Designed and implemented automated error handling and retry mechanism on transient
system failures, achieving 99.99% service availability and curbed the service error rate
to less than 0.05%.
- Achieved 13% reduction in Spring Boot service response time by introducing cache
replication and persistence with EHCache and Redis.
- Mentored three graduate hires on project.
Full Stack Engineer Intern @ Stackroute
June - November 2019
-
Built a pluggable help desk system using IBM Watson that decreased response time to the
customer’s queries by 20% by generating automated response, routing customer to right
agent for one-to-one chat and displaying relevant statistics on dashboards.
-
Developed several re-usable NPM packages using Angular 8, Redux, Chart.js for front-end
components, Node.js and Express for backend services, unit tested with Karma tests. The
packages went on to reduce development time for several key projects by 20%.
ML Intern @ Center for Development of Advanced Computing (CDAC)
Open-Source QnA Transformer-based Chatbot
-
Built a Chatbot with Transformers for generating answers to questions specific to
PyTorch, employing two pretrained HuggingFace BERTs as encoders, FAISS for context
matching and retrieval and pretrained BART as decoder.
-
Collected over 10k data points from official Pytorch documentation, forums, and video
captions by building a web scrapper on Python.
Open-Source Object, Depth and Plane Detecting Python API
-
Combined 3 CNN architectures to build a REST service capable of detecting PPE objects
such as masks, hardhat, vest, boots (YOLOv3), estimating monocular depth (Intel MiDaS)
and plane segments (NVIDIA PlaneRCNN) from an image with a cumulative accuracy
of 86%.
