The smallest discrepancy exposed the larger assumption

The discrepancy was one hundredth in a synthetic example, but the disagreement was about the rule rather than numerical precision. A shared component rounded after adding line amounts; the approved client rule rounded each line first. I had supported the component because it removed repeated implementation and appeared to make the system more consistent. It had also made a business choice that belonged to the client. The example changed how I ask engineers to explain an abstraction: which differences does it preserve, and which does it silently decide are unimportant?

I grew up in Palakkad and completed computer engineering at the fictional Merqovi Engineering College in June 2000. In July I joined Vaqzeri Industrial Software as a software engineer. The company employed eighty people and recorded ₹11 crore revenue for FY ending March 2001. I worked in a six-person development group with no direct reports or spending authority. Our software represented industrial processes defined by the customer. Senior engineers and qualified client staff owned acceptance. My responsibility was to make behaviour reproducible and understand why a result differed from the stated expectation.

I became engineering lead in June 2006. Vaqzeri had 170 employees and ₹38 crore revenue for FY ending March 2007. Twelve colleagues worked through two leads within ₹1.1 crore annual engineering expense. The appointment moved me from individual components to the relationships between them. A shared library could reduce duplication while introducing a common assumption into systems whose customers needed different behaviour. I learned to ask for examples at the boundary of the specification, including cases that looked almost identical but required different outcomes. I also learned that explaining a difference could take longer than correcting the code once the difference was understood.

In September 2012 I joined Qelvazi as platform architect. The supplier had ninety employees and ₹36 crore revenue for FY ending March 2013. I led eighteen engineering employees through three leads, with four contractors separately counted, within ₹2.6 crore annual operating expense. Insurance introduced rules whose authority came from the client insurer, not from the software supplier's preference. A configurable product could make implementation more flexible without making every proposed configuration approved for business use. I needed to preserve those boundaries while helping clients obtain reliable behaviour from a common platform.

Promotion to engineering director followed in July 2017. Qelvazi had 170 employees and ₹92 crore revenue for FY ending March 2018. My team comprised 36 employees through four leads and six contractors, within ₹7.2 crore annual operating expense. I became responsible for the evidence behind release decisions across several clients. That period made configuration inheritance, diagnostic detail and reversal more practical concerns. A convenient default could cross a client boundary, a helpful trace could contain more detail than troubleshooting required, and returning to an older application did not necessarily restore a compatible operating state.

I became CTO in January 2022. Today I lead 57 employees through six leads, with eight contractors counted separately. My FY2026–27 authority includes ₹12.8 crore operating expense and ₹3.6 crore approved change investment; finance determines the accounting classification. Qelvazi employs 260 people, serves 22 insurer clients and earned ₹148 crore revenue for FY ending March 2026. Those are supplier figures, not insurance premium. I report to Joseph Belzethi. Reema Navqeri retains engineering acceptance, and client insurers retain their product, pricing, coverage and claims decisions.

The calculation review found 37 differences in 500 synthetic fixtures. We spent ₹54 lakh on the correction and added ₹12 lakh annual support expense; all 500 subsequently matched the approved map. The configuration review found 23 incorrect inherited defaults in 180 synthetic scenarios. Explicit client-scoped configuration cost ₹88 lakh and added eight milliseconds to the 95th-percentile read latency under the stated benchmark. Those were controlled engineering findings. I did not turn them into claims about actual customer loss, affected policies or proven incidents that the tests had not established.

Diagnostic traces presented a different tradeoff. In 960 synthetic trace events, 146 contained specified test-policy fields that were unnecessary for the chosen diagnostic purpose. The revised structured trace passed a later 1,440-event review without those fields, but median troubleshooting time across twelve defined rehearsals increased from eighteen to twenty-seven minutes. The ₹39 lakh programme made us accept a less convenient diagnostic account. The rollback programme likewise retained a cost: ₹1.1 crore of work, a four-week delay and two features still deferred after 38 of forty compatibility scenarios passed.

I have no statutory directorship and have not yet registered in the IICA independent directors databank. Twenty days is a provisional allowance for one external role, subject to preparation, consent and the actual demands of a mandate. My Malayalam, Hindi and English support my working context without implying every client interaction occurs in the same language. I would bring experience of a supplier's commitments and the discipline to ask what an engineering result establishes. I would also need to learn the proposed company's economics and obligations, so that a useful technical question becomes a useful board question rather than an invitation for me to manage another team's code.

Open to Board appointments

Languages / Malayalam · Hindi · English