Context and problem

At an energy-sector company, project management tickets and their related context lived in JIRA, and the information was hard to use quickly: search was manual and answers required a lot of expert handwork. They needed a solution that understands the content and history of the tickets and supports search and answer generation through automation. The client had a clear vision of the need and the implementation, which made fast progress possible.

What was done

I built a Retrieval-Augmented Generation agent for the JIRA data. Tickets were indexed into a searchable form, and the agent combined retrieval with LLM-based answer generation so that the user received a ready answer and relevant context without having to read through long threads by hand.

The implementation was done in an Azure environment: the RAG assistant ran as Python in Azure cloud functions, and the assistant was integrated into JIRA as a Java application. The infrastructure was written as Bicep code, which made fast integration into different environments possible. The data pipeline from JIRA into the document index was implemented with the existing Python JIRA package and cloud functions. I was responsible for the technical design and implementation of the project. Implementation and delivery February–March 2025.

Key technologies: RAG, LLMs, Azure, Python, Java, IaC (Bicep), NLP.

Outcome

The agent enabled intelligent search and automated answer generation on top of project management data. The solution was taken into production once completed. The project team agreed to revisit further development as needs become clearer — working production first, expansion as needed. The solution showed that a structured but text-heavy source like JIRA can be put to use with a RAG agent without a heavy data engineering project.

← Back to assignments