Shlok Pandey supervised by Prof. Veera Prakash Yalla received his Master of Science in Computer Science & Engineering (LCD) by Research. Here’s a summary of his research work on Accelerating the First Commit via IssuePilot: An LLM-Driven Framework for Enterprise and Open-Source Onboarding
Open source softwares rely heavily on new contributors due to the continuous influx of new individuals interested in software and trying to get started with it. However, it often becomes difficult for newcomers to find issues that match their interests, which makes the onboarding process a significant hurdle. There do exist beneficial practices like using labels such as “Good First Issue” as an attempt to guide newcomers; however, it fails to account for the developer’s specific skills, background, and current learning interests. Even if an individual selects a task (issue) to work on, it is difficult for them to figure out where to get started within the codebase.
Therefore, this thesis proposes IssuePilot, an agentic system tool that generates personalized issue recommendations for the repositories of interest, along with personalized onboarding documentation. The platform extracts developer profiles using uploaded resumes and interactive QnA sessions, creating a compressed profile of their skills and interests. Then an Artificial Intelligence (AI) based recommendation agent ranks open issues across the repositories in order to find the best match for the developer. Once the issue is selected and assigned, the system automatically analyzes the repository’s code structure and generates personalized onboarding documentation. This documentation gives the developer the necessary context to start working on the issue.
This system is evaluated using a UTAUT2 based questionnaire. These questions cover profiling accuracy, recommendation quality, system architecture, and onboarding documentation. The findings show that IssuePilot was able to capture developer interests accurately and outperforms generic resume-only matching. This tool is then scaled to IssuePilot Enterprise, an onboarding platform tailored specifically for enterprises. This thesis also mentions the scalability challenges and introduces a full-fledged software system. This system follows Role Based Access Control and makes use of a resilient Large Language Model (LLM) based gateway. Several approaches to tackle issues such as reducing non-deterministic outputs from LLMs are also discussed. At the end, a testing architecture is setup to validate the system.
Ultimately, this work demonstrates that agent-based developer profiling and issue matching are viable at an organizational scale, making onboarding more efficient. The aim of this system is to allow the teams to spend less time on the administrative triage and more on what they are best at, i.e., writing impactful code.
August 2026