Work+Research

Research

ELASTIC: Evaluation of Large Language Models and Abstract Syntax Tree Parsing for Identification of Code Smell Detection

Building on SmeLLM, this approach uses Static Analysis (particularly Abstract Syntax Tree parsing) to create rules to detect code smells.

Advised by Dr. Joydeep Mitra

Northeastern University

SmeLLM

A code smell is defined as an anti-pattern when writing code. Presence of code smells can make code harder to read, more difficult to maintain and generally harder to debug or make changes to. The increasing usage of LLMs in the field of Software Engineering lead us to examine the effectiveness of popular LLMs in identifying these code smells and developing a tool that can help evaluate the effectiveness of these LLMs and provide the user feedback about their code.

Advised by Dr. Joydeep Mitra

Northeastern University

Enhanced Location Entity Recognition and Extraction

Enhancing the CLAVIN-NERD model to achieve more accurate location entity recognition and extraction capabilities. Exploring different approaches, such as retraining, fine-tuning and post processing-procedures to improve the model's performance.

University of Auckland

Work

Oneture Technologies

Software Engineering Intern

Jan - July 2024

  • Created a UDP Packet Streamer for ICICI Securities in C# using .NET 6 with a GUI using WinForms to upgrade a legacy .NET 3.5 application
  • Increased packet throughput by 10x using multi-threading and data structure optimization
  • Introduced a resilient monitoring tool that enhanced packet visibility and a comprehensive logging solution that captured and analyzed over 100,000 packets daily.
  • Develop an application using React, Tailwind CSS for an intuitive and responsive front-end, and Node.js with Express for efficient back-end functionality with a MySQL Database using advanced queries.

Oneture Technologies

Software Engineering Intern

Oct - Dec 2021

  • Implemented an automated pipeline to create documentation for ICICI Securities’ iBreeze API to eliminate manual documentation
  • Created using static site generator Slatedocs for documentation and Node.js along with Shell (Bash) and Postman SDK for a lean server infrastructure
  • Developed a Web CMS to manage the generated documentation using EJS and Node.js to modify any resource