Software testing has always been the quiet guardian of software quality. But for too long, the way we test has lagged behind the way we build. In a new paper published in the International Journal of Advanced Research in Engineering and Technology (IJARET), Mr. Yerram makes the case that this gap is finally closing, and that the shift underway is structural, not just another round of incremental tooling. It is the kind of paper that does not just describe where the field is, but quietly redraws the map of where it should go next.
The paper, Modernizing Software Testing: AI, Telemetry, and Agentic Quality Engineering, introduces a discipline Mr. Yerram calls Agentic Quality Engineering. The idea is straightforward but ambitious: autonomous AI agents generate, run, prioritize, and repair tests on their own, while continuously learning from live production telemetry. What you end up with is a feedback loop that ties pre-release testing to what actually happens once software is in users’ hands, two things that have historically operated in near-total isolation. It is a deceptively simple reframing, and one that few in the field have articulated with this much clarity or conviction.
Anyone who has watched a minor UI change shatter hundreds of brittle test scripts will recognize the problem he’s describing. Mr. Yerram characterizes conventional testing as fundamentally reactive. It covers the scenarios humans thought to check, drifts away from real usage patterns, and demands constant maintenance just to stay functional. His framework attacks each of those weaknesses through five connected stages: intelligent test generation, adaptive execution, self-healing maintenance, telemetry-informed risk analysis, and autonomous triage that learns over time. What impresses here is not any single stage in isolation, but the way he assembles them into a coherent whole. Where most of the field has been chipping away at individual problems, Mr. Yerram has the architectural instinct to see the entire system at once.
The figures he reports from surveyed enterprise rollouts are the kind that make engineering leaders pay attention. Teams adopting these methods saw test maintenance overhead drop by as much as 70 percent, defect detection efficiency improve by roughly 45 percent, and mean time to detection for production-impacting defects fall by about half. To his credit, Mr. Yerram resists the temptation to oversell. A good portion of the paper deals with the messier side of autonomy: non-determinism, over-trust in plausible-but-wrong agent decisions, security exposure, and the question of who is accountable when an agent gets it wrong. His answer is calibrated autonomy, where agents handle the repetitive, high-volume work and humans keep authority over strategy and release decisions. That intellectual honesty is rare, and it is precisely what separates a serious contribution from a hype cycle.
The most useful part for practitioners may be the four-stage maturity model, which lays out a realistic path from AI-assisted testing to fully continuous agentic quality engineering. It reflects something a lot of vendor pitches conveniently ignore, which is that enterprise transformation almost never happens in a single leap. By handing teams a pragmatic, staged roadmap rather than a finished ideal, Mr. Yerram does something genuinely generous: he makes his vision adoptable.
What gives the work weight is its balance. It is genuinely optimistic about where AI can take software quality, but it stays grounded in the governance realities engineering teams actually deal with. As systems get more complex and release cycles keep accelerating, Mr. Yerram’s picture of testing as a living, self-improving, telemetry-aware loop reads less like speculation and more like a preview of where the field is going. Contributions like this one do not come around often. They tend to be remembered as the moment a discipline started thinking differently about itself, and Mr. Yerram has positioned himself firmly at the front of that conversation.
About the Voice Behind the Work
Mr. Yerram Sai Rakshit is a Technology Leader at Visa, Inc., in the United States, with prior tenure at Cox Automotive and Teladoc Health Inc., working at the intersection of software quality engineering, automation, and applied AI. He has spent his career building reliable systems at enterprise scale, and his work centers on modernizing how software quality is engineered and assured by bringing autonomous agents, observability, and continuous learning into the testing life cycle. A clear and forward-thinking voice in the field, he continues to argue for a future where quality engineering is intelligent, adaptive, and demonstrably trustworthy, and his latest work marks him as one of the practitioners genuinely shaping that future.
Read the Full Publication
Abstract: https://iaeme.com/Home/article_id/IJARET_17_03_002
Full Text: https://iaeme.com/MasterAdmin/Journal_uploads/IJARET/VOLUME_17_ISSUE_3/IJARET_17_03_002.pdf
DOI: https://doi.org/10.34218/IJARET_17_03_002