Nikhil Agarwal

Experience

• 

  MAQ Software

  May 2015 - Jun 2015

  Software Engineering Intern

  Worked on automating testing of web reports portal that MAQ maintains for user data of different clients using Selenium.

• 

  Microsoft

  May 2016 - Jun 2016

  Software Developer Intern

  I created a windows form application to automate the process of structured data extraction. I built a framework to extract XPaths from HTML DOMs of data sources (for example Wikipedia and Wikia) which were then used to extend to similar web pages to extract data. Usecases involved data regarding celebrities, fictional characters, and events among many more. This was a dev tool aimed at aiding the Bing experiences team.

• 

  Oracle

  Jun 2017 - Jul 2019

  Applications Engineer

  Contributed to 100+ enhancements in the fusion supply chain management tool relating to inventory, costing, and accounting including functional improvements, user interface additions, SQL query additions and improvements, and designing JUnit tests.

• 

  Goldman Sachs

  Jul 2019 - Apr 2021

  As part of the compliance division of Goldman Sachs, I facilitated the team in detecting deceptive, malicious and fraudulent transactions. My role required me to look at the technical aspects of this task like data processing and building business logic. The key area of focus was to create innovative solutions to make it faster and easier to run a wide variety of such surveillance on financial data. My primary work involved building an end to end product to extract trade orders and executions with high risk concerning violation of particular surveillances and trade norms.

  • Associate

    Dec 2020 - Apr 2021

  • Senior Analyst

    Jul 2019 - Dec 2020
• 

  Manning College of Information and Computer Sciences, UMass Amherst

  Aug 2021 - Jun 2022

  Graduate Student Researcher

  Working under the guidance of a Doctoral candidate in Computer Science department, UMass Amherst. Working on using multi facet softmax instead of a single softmax to predict/recommend the next item a user is likely to buy based on user history. Experimenting with different recommendation models like SASRec and GRU4Rec on top of different datasets like Amazon book reviews, movielens dataset for improvement in metrics like hit@10 and recall@10.

• 

  Google

  Jan 2022 - May 2022

  Graduate Student Researcher

  As a part of an independent study, collaborated with researchers from Google, Purdue university, and UMass Amherst. Experimented with specific designs of Multi layer perceptrons and tested their performance on different tasks like set summation and image classification against architectures like CNN, aiming to design unbiased architectures that implicitly learn data symmetries.

• 

  Meta

  May 2022 - Aug 2022

  Software Engineer Intern

  On-device deployment of BERT for Question Answering: Worked on deploying quantized BERT language model on Meta Quest 2 (Oculus). The project included conversion of the pytorch model to a mobile model, followed by loading and inferring the model in C++.The final step of the project was to deploy the C++ binary on Oculus and reporting improvements in runtime and memory usage compared to the full precision model.

• 

  Amazon

  Jan 2023 - now

  Software Development Engineer

  Started working for AWS Marketplace in the SaaS buyer experience team.- Migrated fetching the pricing information to a compliant API on the procurements page.- Reduced the page load time for product details page and the procurements page by caching the response of the API used for fetching pricing information.-Modified the UI/UX to show alerts related to exceptions on the web application.-Currently working on enabling multiple currencies for offers. Made modifications in the backend data models to accommodate the currency parameter.-Enabled lazy loading of available offers in the dropdown on the procurement page to reduce worst case page load latency to 0.1x Show less