Challenge: Improving the alignment of source code repositories with project management systems

Solutions: An innovative platform powered by generative AI that automates workflows for instant updates between code repositories and requirements tools, streamlines tasks, and provides real-time insights to mitigate risks

Benefits:

• Empowers teams to quickly assess project progress
• Improves code quality
• Reduces expenditure of time and effort
• Heightens collaboration and efficiency
• Provides faster development
• Helps identify potential risks
• Ensures alignment with project goals

Key performance indicators

• 40-50% faster approval time thanks to AI answers to questions
• 30-40% faster identification of code sections for enhancement
• 50-60% time savings in tracking alignment between requirements and code
• 50-60 % fewer bugs thanks to upfront AI identification of gaps in coding

The goal of this platform was to alert project management when code is introduced or modified, providing instant updates between the product management and engineering teams. Powered by generative AI, this innovative platform automates workflows, aligns tasks and empowers teams to make smarter decisions in real time. Because project managers have a dashboard where they can see risks and progress instantly, the platform helps teams deliver better products faster and with fewer difficulties. In addition, the system automatically checks whether pull requests (changes to the code submitted for review by developers) match requirements, giving intelligent feedback to help project managers make informed decisions and speed up product delivery. And because it works with existing tools, like GIT and JIRA, there’s no need for big changes to workflows.

When developers make pull requests, the AI updates the requirements system, applying advanced large language models (LLMs, such as such as Sentence Transformer, BERT, Llama 2 and Code Llama) as well as natural language processing (NLP) techniques to analyze the linkages between requirements and source code; it suggests improvements, ensuring that the code responds to real needs and objectives. The search and retrieval system uses vector similarity (QdrantDB and ChromaDB) to improve performance. Amazon Web Services (AWS) provides hosting to ensure that the platform can readily handle large projects.

ALTEN’s expertise proved invaluable in deploying and fine-tuning these tools. Clear roles and responsibilities were assigned for each member of the team, which included one architect, two backend developers, one frontend developer, one AI engineer, one UX designer, two DevOps engineers, and one QA specialist. Collaboration across roles was seamless, facilitated by tools like Slack, Jira and GitHub. Demo-able sprints every two weeks allowed for regular feedback and kept development on track. Testing, CI/CD pipelines and proactive communication ensured tasks were completed without delays.

The number of bugs reported in production was minimal, showcasing the strength of the development and quality assurance processes. Thorough testing enabled quality gates at every phase of the project. Regular code reviews and adherence to coding standards maximized productivity throughout, reducing the chance of defects making it to production.

There are numerous benefits to the new platform. By synchronizing code with requirements, the AI system bridges the gap between product and engineering teams, ensuring that everyone works toward the same goals. In addition to checking whether the code fulfills requirements, it breaks it down into smaller tasks and updates task status automatically. Thanks to up-to-date insights on progress, project teams can address potential issues before they escalate. Project managers can ask questions and receive responses before merging, helping them to make better decisions. The accuracy of the AI-generated feedback makes the development process more efficient and minimizes the room for error. Another advantage is that the automated workflows reduce manual work, freeing up developers and managers to focus on high-value tasks. And because the platform works with familiar tools, it minimizes the learning curve. A strong backend – with application programming interfaces for authentication, dashboards, background tasks, and AI integrations – together with code designed to handle future scaling, make the project efficient and easy to maintain, despite its complexity.