More responsibility for the meaning of a test

Mandate

Make software behaviour under difficult operating conditions a funded, testable part of product engineering.

Outcomes

• Funded ₹48 lakh update redesign and recorded the client's ₹180 per-device memory consequence
• Kept an independent message holdout in the parser acceptance evidence
• Agreed record identity across repeated deliveries with the receiving-system owner
• Selected a compression compromise using both byte volume and processing cost

Decisions in context

My CTO appointment brought the engineering account directly into discussion with the chief executive and, when invited, the board. I lead five heads rather than personally choosing every implementation. The question I want each team to answer is how its evidence relates to the behaviour the customer expects. Our four recent cases show different answers: a recovery path after interrupted power, another challenge after a parser repair, record identity despite repeated transport and an optimisation that moves cost into processing. The annual expense and laboratory investment are separate responsibilities within a 520-person company. I am not a director and do not claim its revenue as my P&L. I also preserve the client's authority over final product acceptance, because a software supplier's confidence must not silently become approval of an entire device or its intended use.

Reason for the move

Current executive role; considering one external oversight opportunity after preparation and clearance.

Mandate

Turn architecture and validation knowledge into repeatable engineering decisions across growing project teams.

Outcomes

• Required project plans to identify significant untested conditions
• Connected validation effort with the consequence of a software failure
• Made handover limits and client acceptance responsibilities explicit

Decisions in context

Managing eighty-two engineers changed my relationship with a test report. I could no longer follow every experiment, and a summary that omitted its conditions could travel much further than the engineer intended. I worked through three managers to make the unresolved question visible in project reviews, alongside the work needed to answer it. That did not mean adding every imaginable test to every schedule. It meant selecting evidence that related to the promised use and explaining the consequence of leaving a condition unexplored. The separate equipment programme also required a view of what capability a laboratory purchase would create. I wanted teams to explain why a new instrument or fixture was needed before assuming that more equipment automatically produced better validation. The next appointment extended that investment judgement across Pelvori's engineering direction.

Reason for the move

Promoted to CTO with responsibility for company-wide engineering direction and investment.

Mandate

Clarify device-software interfaces, expected error behaviour and the evidence supporting engineering handover.

Outcomes

• Documented assumptions about retries and incomplete device transitions
• Separated software test results from whole-product approval
• Brought receiving-system owners into interface reviews

Decisions in context

Joining Pelvori gave me responsibility for architecture across more projects than I could hold in my own working memory. I learned to make an assumption useful to another team: what the sender believes, what the receiver confirms and what remains uncertain after a restart. The twenty-four-person group combined architecture and validation so that a design decision could be challenged while it was still practical to change. Our clients retained authority over the complete product, a boundary I learned to state carefully in handover discussions. A clean software test was valuable evidence but did not settle hardware, deployment or wider acceptance questions. I also became more attentive to the commercial consequences of a technically sensible change. The promotion in 2016 asked me to carry those disciplines into project management and the allocation of a larger engineering budget.

Reason for the move

Promoted to lead engineering as the business expanded its client programmes.

Mandate

Deliver client firmware while making interface assumptions and difficult operating conditions part of the work plan.

Outcomes

• Required explicit agreement on retry and error-handling behaviour
• Explained validation time in client delivery discussions
• Reviewed reuse decisions against each device's actual constraints

Decisions in context

A services business exposed me to the danger of recognising a familiar interface too quickly. Two devices could use similar messages while expecting different behaviour when the connection failed or a response arrived late. Leading eleven engineers meant I had to make those differences discussable before a reusable solution was promised. I also learned to explain why a test that looked unusual to a customer might be essential to the service behaviour it expected. Our ₹1.8 crore team expense did not give me authority over the employer's full commercial result, but it did require me to account for engineering time. I left to take a wider architecture role, carrying the view that a clear assumption at the start of a project is often more valuable than a clever recovery from an avoidable disagreement near its end.

Reason for the move

Joined Pelvori to work on a broader embedded architecture and validation remit.

Mandate

Develop and test firmware for settings, inputs and communications in industrial monitoring accessories.

Outcomes

• Learned to test interrupted operations rather than only normal sequences
• Recorded recovery behaviour with the development team
• Connected bench observations with the support conditions of installed devices

Decisions in context

I began with the ordinary work of making a small device remember a setting and communicate a result. The development group had six people, so a problem was quickly visible and usually easy to discuss around the bench. That convenience could conceal the customer's reality. A unit in use would not always have stable power, a perfect connection or the engineer who wrote the code nearby. An interrupted demonstration made me take those conditions seriously. I had no purchasing authority and could not change the product specification by myself; I had to explain the observation to the development manager and help test an agreed response. The move to Heshvani offered a first team-lead role and exposure to a wider range of devices, while retaining the practical discipline I had learned from those early bench failures.

Reason for the move

Moved into firmware services to work across more equipment designs and lead a small team.