Chief Technology Officer / India
A more complete picture was not always a better answer
- Technology team through five leads
- 61 staff
- ₹ 12.2 crore FY 27 expense; development programme separate
- Survey processing, software and client-side delivery
- 31 years
- Career beginsJuly 1995; CTO sinceApril 2022
- Provisional annual board capacity
- 25 days
- Employer consent, relationships and calendar still to examine
Decisions & consequences
01Withdrawing the positional claim behind a detailed image
In 2022 I endorsed a proposal that treated three-centimetre ground sample distance as evidence of three-centimetre positional accuracy. The presentation made the imagery look impressively detailed. Arun Jezmavi asked what independent comparison supported the separate claim about position. Twenty-four checkpoints, withheld from the processing fit, were examined against the practice's predefined criterion. Seventeen results were within that criterion and seven were outside. The examination did not support the statement I had allowed into the proposal.
The commercial sponsor wanted us to keep the promised release and describe the image specification more prominently. That would have preserved an attractive number while leaving its interpretation unresolved. I supported withdrawal of the positional claim and ₹ 68 lakh of investigation and further acquisition work. The qualified survey practice retained responsibility for method, interpretation and acceptance. We did not replace its criterion with a more convenient one or suggest that the seventeen passing results made the other seven immaterial.
Release remained held for the relevant acceptance decision. This account asserts no corrected accuracy result and no universal certification from a set of twenty-four checkpoints. The finding concerned the defined project examination. Our processing team preserved the comparison material and helped investigate the discrepancy; the additional work still needed its own evidence before a revised performance statement could be made. The expense was a cost of addressing the unsupported commitment, not a measured return on investment.
I had known the conceptual distinction but failed to protect it when the commercial proposition compressed several properties into one number. That made the correction my responsibility as well as a technical task. I would now have the survey owner review the wording of the performance claim before it becomes part of a delivery commitment. A detailed product can be valuable without being suitable for every positional use. The client needs to understand the supported use and the remaining uncertainty before appearance becomes an informal assurance that neither the proposal nor the work can substantiate.
Keeping ten elevations outside the reconciled height account
In 2023 a client-facing integration assembled nine hundred height records. Seven hundred and eighty carried the reference basis required for the common account. One hundred and twenty came with a different or insufficiently established basis. The delivery team proposed aligning the latter group by applying a convenient constant adjustment. The resulting column would have looked coherent, but the proposed operation did not establish that its values represented comparable heights.
Arun asked for source examination and an appropriate qualified reconciliation instead of an assumed shift. I supported ₹ 47 lakh for the review and supporting technology work. The survey practice determined the methods and evidence needed for each questioned source. My team retained the original values, reference descriptions and authorised changes so that reconciliation did not erase the path back to what had been supplied. A completed transformation in the software was not itself the acceptance decision.
Qualified review reconciled 110 of the questioned records. Ten remained held, leaving 890 supported records within the defined common account. Those figures partition the same nine-hundred-record population; they are not an estimate of all future source quality. We did not claim that height reconciliation approved a drainage design, settled a land question or established the accuracy of every underlying observation. Those uses required their own responsible professionals and evidence.
The client sponsor was concerned that an incomplete set would postpone work. That concern deserved a practical answer about which records could be used and where alternatives were required, rather than a polished table that concealed the gap. We delivered the supported set with the ten exceptions identified for their owners. I would now establish the required reference information before accepting a source, with an explicit response when it is absent. Technology's contribution was to preserve meaning through the process and make unfinished examination usable for the next decision. It was not to manufacture compatibility because an interface could perform an arithmetic operation.
Leaving three hectares unobserved after further field access
A 2024 proposal described a sixty-hectare image footprint as though it supplied a measured ground surface throughout the area. Examination showed forty-eight hectares of observed ground and twelve obscured. The visual coverage and the ground evidence were different. A delivery lead proposed filling the obscured areas through interpolation and presenting a continuous surface, arguing that a visibly incomplete product would be difficult to explain to the client.
I supported ₹ 59 lakh for additional authorised field work and the associated processing. Field permissions and access decisions remained with the proper project and land owners; neither a technology budget nor a delivery date entitled us to enter restricted areas. The survey practice determined what additional observation would be useful. My team kept the distinction between observed ground, any clearly identified estimate and areas without a supported answer in the delivery representation.
Further work provided observations for nine of the twelve hectares. Three remained inaccessible. The final coverage account therefore contained fifty-seven hectares of observed ground and three hectares without observation. We did not call the remaining area measured, infer that vegetation had been penetrated everywhere or claim that the additional observations approved a construction decision. The result described evidence coverage within the defined footprint, with the ordinary survey and engineering acceptance responsibilities still applicable.
Keeping a gap required the commercial team to discuss a limitation instead of presenting a complete-looking surface. It also gave the client a better basis for deciding whether to change the proposed use, seek further authorised access or retain the uncertainty. Interpolation may be useful when its purpose and assumptions are understood; the objection was to relabelling it as observation. I would now ask the proposal team to distinguish image footprint from usable ground evidence before agreeing the scope. A blank area can be an important part of a truthful spatial product when filling it would communicate knowledge the service has not obtained.
Separating usable source files from permission to deliver them
In 2025 a combined spatial product depended on twelve separately licensed source datasets. A processing demonstration showed that the files could be combined successfully. The sales team interpreted that technical result as readiness to supply the resulting client deliverable. Review of the actual source terms found seven that expressly permitted the intended delivery, two that restricted it and three whose scope needed clarification. File access had not settled the permission question.
We allocated ₹ 31 lakh to qualified review and procurement of possible replacement material. The commercial sponsor preferred to proceed with the combined demonstration while seeking clarification later. I supported keeping the restricted and unclear sources outside the proposed client delivery until their position was resolved. The rights review belonged to the appropriate commercial and legal owners; technology recorded source identity and processing dependencies so their decisions could be applied to the product that would actually leave the business.
The reported outcome remains seven expressly permitted sources, two restricted and three requiring clarification for the intended use. This account does not assert that the replacement procurement resolved every issue or that combining a dataset necessarily removes its restrictions. We retained the unresolved part in the delivery plan. A change in client, format or proposed onward use could require a different review rather than inherit the conclusion reached for this particular proposition.
The decision created a narrower immediate offering and additional expense. It also made the commercial team confront a dependency that a successful technical demonstration had obscured. I would now include the intended delivery and use when source material is selected, so a later product promise does not outrun the permissions supporting it. The executive lesson is about coordination: technical usability, professional fitness and contractual permission answer different questions. A CTO must make those questions visible to their owners and resist treating the part his team can demonstrate as evidence that every other condition has been satisfied.