Varun Mukka is an Engineering Architect with a focus on Infrastructure Quality. Since that time he has become a leader in seeing how artificial intelligence (AI) and machine learning (ML) can be integrated into test automation and continuous verification practices. His groundbreaking work at Okta has established a new best practice for speeding up deployment cycles, all the while minimizing post-release failures. Mukka’s methodology conforms directly to industry best practices, and their focus on ongoing testing and verification are key.
Mukka’s journey started with the data cleansing process of compiling several years’ worth of test results and production incident logs. This important, groundwork-laying effort set the stage for his team to deploy AI-powered solutions more strategically. As the integration of AI took root, the deployment frequency at Okta increased significantly, showcasing the potential of innovative practices in the field of infrastructure quality.
The Integration of AI in Workflow
With AI fully integrated into Okta’s workflow, that was a watershed moment for Mukka and his team. They put off continuous verification by leveraging testing hooks within production and staging environments. In order to meet the needs of their customers, they greatly increased their deployment cycle. This bold step didn’t just make their process more efficient, it helped them become more reliable overall.
Mukka added that after the new practices got established, the deployment frequency saw a significant jump. He stated, “Since adopting ML in our quality process, our key quality metrics have all trended upward.” This increasing trend is a testament to the power of AI-driven testing paired with continuous verification to improve infrastructure quality.
Mukka further implemented ML algorithms to track metrics after deployments to identify outliers. This forward-thinking stance helps teams to catch problems in their pipeline before they hit production, reducing the risks of post-release issues.
Challenges and Solutions in AI Adoption
Even with these clear benefits, the journey to integrate AI into Okta’s quality processes was not without challenges for Mukka. Perhaps the biggest challenge was making sure that the AI models had quality data to learn from to model accurately. Mukka emphasized the need for ongoing training and fine-tuning of these models, likening the process to “raising a child, requiring patience and good guidance.”
Additionally, fear within the team about AI’s ability to replace their jobs was widespread. Mukka tackled these concerns directly by clarifying AI’s recommendations and engaging team members in interpreting the outcomes. He noted, “We tackled this by treating the AI as an assistant, not an oracle.” This collaborative approach helped build a culture of transparency and acceptance among agency workers regarding the use of AI.
Mukka began modestly by adding AI-powered checks first where legacy scripts had the most difficulty. “When those early projects caught critical bugs that we would have missed otherwise, it validated our approach,” he remarked. This approach enabled his team to introduce new technologies in a way that wasn’t too jarring.
The Future of Continuous Verification
Looking forward though, Mukka sees a day when ongoing verification will develop into a more intelligent and autonomous process. He hopes that these developments will result in self-adaptive systems that can automatically and efficiently manage many different quality processes.
Mukka further remedies this misconception by declaring efficiency is not just about how fast you go. He emphasizes the importance of smarter resource utilization, stating, “Engineers will focus more on defining quality goals, interpreting AI findings, and handling the creative and complex aspects of quality that AI can’t easily grasp.” This dynamic vision underscores all the importance of human intuition and creativity in test design and analysis. Even as AI makes these processes better, our instincts are still invaluable.
“AI can augment your testing, but it won’t replace good test design and analysis,” he added. “Use AI to handle scale and complexity, but always validate its findings with your expertise.” This balanced approach reflects Mukka’s resolve to harness the power of technology, while still keeping the priceless knowledge of human professionals at work.
