Developers often search across GitHub issues, pull requests, release notes, documentation, and source files to understand why an error happened and how it was fixed. Traditional code search can find matching text, but it usually does not connect the full troubleshooting trail: the original error report, related issue discussion, fixing pull request, changed files, release version, and updated documentation.
In this session, the speakers will show how to build a GitHub troubleshooting graph with Neo4j. They will demonstrate how public GitHub repository data can be modeled as a graph of Issues, Pull Requests, Files, Releases, Comments, Docs, Error Signatures, and Fix Paths. The session will show how Neo4j can combine semantic search with Cypher traversal to move from an error message or stack trace to the most likely issue, fix PR, affected files, and supporting evidence.
You will learn how to design the graph schema, ingest public GitHub data, create embeddings for issues and docs, use hybrid retrieval, and write Cypher queries that explain the path from problem to fix. The session will also cover how to evaluate the system using closed issues and linked pull requests as ground truth.
By the end, you will understand how graph-based retrieval can go beyond code search and help developers debug faster, trace fixes, analyze upgrade issues, and build more useful AI coding assistants.
Speaker
Ankush Rastogi
Data Solutions Specialist
I am a Senior Data and AI Engineer with over 10 years of experience building large-scale data platforms and production-grade AI systems. My work spans data engineering, multimodal AI, large language models, and real-time analytics, with a strong focus on enterprise-scale deployments in areas such as contact center analytics, compliance automation, and AI-driven decision systems. I have led the design and implementation of high-throughput data pipelines, GPU-accelerated inference systems, and end-to-end AI platforms that operate reliably in regulated, real-world environments.
In addition to my industry role, I am an active researcher and author, with peer-reviewed publications and ongoing submissions in IEEE and Springer venues. I also serve as a reviewer and judge for international conferences, journals, and hackathons, and regularly mentor within the global data and AI community. My talks typically focus on practical, experience-driven insights such as operating data platforms at scale, bridging data engineering with LLM systems, and designing robust architectures that move beyond experimentation into production.
