In The News
-Sathish Raman
A batch job that misses its window by twenty minutes rarely makes headlines. It shows up instead as a report that lands late, a reconciliation that has to be rerun, an operations team quietly working past its shift to catch up before the next business day starts. Hariprasad Pandian has spent more than seventeen years working on the side of enterprise software where these small delays get caught, measured, and traced back to a cause, across banking, retail, energy, and telecom systems that most customers never think about directly.
Hariprasad Pandian, a performance engineering veteran, reveals his unique approach to preventing critical system failures. Discover how he identifies hidden bottlenecks and optimizes complex enterprise software across diverse industries, ensuring robust performance and avoiding costly downtime. His insights are crucial for anyone seeking unbreakable systems.
Pandian’s title has usually been some version of performance engineer, a role that sits between the developers who write an application and the operations staff who keep it running. In practice, the job means finding out how a system behaves under real load rather than ideal conditions: what happens to response times when a thousand people log in at once, how a database query degrades as a table grows past a certain size, where a batch process starts to slow down as data volumes increase year over year. He has done this work at Citibank, Morgan Stanley, Bank of America, Northern Trust, Lowe’s, Duke Energy, and Zions Bancorporation, moving between Chennai, London, Charlotte, and Salt Lake City over the course of a career that began in late 2006.
One project illustrates the pattern clearly. While working on enterprise infrastructure projects tied to Duke Energy between 2016 and 2017, Pandian’s analysis led to an architectural change that cut the application’s overall processing time by roughly three hundred percent and expanded its capacity to handle one hundred fifty percent more users without a corresponding drop in performance. Getting there required understanding not just where the code was slow, but which constraints (database design, server configuration, data volume, concurrency) were actually driving the slowdown, since fixing the wrong bottleneck can leave a system just as fragile as it started.
A similar discipline shaped his best-documented achievement: a recommendation to change the architectural design of a Big Data Hadoop batch process, cutting its completion time from roughly nine hours to three, a reduction of more than sixty-five percent. In financial and utility environments, batch windows are rarely just a technical detail. Reporting, reconciliation, and downstream analytics often wait on a batch job to finish before the next stage of a business day can begin, so shrinking that window by six hours can change what a team is able to do the same day rather than the next one.
Pandian’s toolkit reflects two decades of accumulated specificity rather than a single favored method. Across his projects he has built and executed test scripts in LoadRunner, JMeter, VSTS, and Silk Performer, monitored production behavior with AppDynamics, Dynatrace, and Wily Introscope, and diagnosed slow queries at the database layer using SQL Profiler, Perfmon, and Oracle Enterprise Manager. At ConsultNet, working on a Zions Bank engagement, he built performance optimization procedures directly into the continuous integration and deployment pipeline rather than treating performance testing as a separate, later step, an approach usually described in the industry as performance-driven development, one he has carried into his current role as senior performance engineer at Zions Bancorporation in Salt Lake City.
What connects a bank’s transaction processing to a utility’s batch reporting to a retailer’s tax exemption system is less obvious than it sounds, since each runs on different infrastructure, under different regulatory pressure, at a different scale. But the underlying failure modes repeat: a resource that runs out sooner than expected, a process that degrades gracefully until it suddenly doesn’t, a system that was tested for the load it was designed for and never for the load it eventually received. Pandian’s work has largely consisted of noticing these patterns earlier than the systems themselves reveal them, then translating that observation into an architectural change someone else can implement.
He holds a bachelor’s degree in electronics and communication engineering from Anna University and a master’s degree in computer science from Campbellsville University in Kentucky, completed between 2020 and 2022 while continuing to work full time. The gap of nearly two decades between the two degrees mirrors the arc of his career: an engineering foundation built early, followed by years spent inside production systems learning what breaks and why, before formalizing that experience academically. It is a sequence more common in performance engineering than in most software specialties, where the most useful knowledge often comes from watching a system fail in a way no test plan anticipated, and then making sure it doesn’t happen the same way twice.
