Make the model answer the real decision

In 2019 I approved a forecast demonstration that appeared stronger than the evidence available at the time could support. A review of 120 historical forecast windows found that eighty-six had used weather observations unavailable at the moment the forecast was supposed to have been issued. Thirty-four used inputs available at that time. The problem was not whether the historical observations were accurate. They answered a later, easier question. I withdrew the performance claim and funded a reconstruction of the evaluation around what the forecaster could actually have known.

I now lead technology at Drevlqari Energy Software in Visakhapatnam. Five heads lead fifty-two research and engineering employees within ₹9.8 crore annual operating expense, with a distinct ₹1.7 crore approved equipment and laboratory capital programme. The company employs 164 people and earned ₹82 crore in FY2025–26, divided between software and implementation or analytical services. I report to CEO Anita Vezlqerni and hold a 0.25 percent beneficial interest through the employee vehicle, without a statutory seat. Our products support industrial energy planning. They do not autonomously dispatch equipment or replace customers’ engineering, safety and commercial approvals.

Later work challenged recommendations that violated a stated planning constraint and hourly summaries that concealed a shorter demand peak. Those cases required the software to admit when an attractive answer did not satisfy the question. I would consider a CTO mandate where evaluation, feasibility and time resolution receive the same attention as the demonstration. An outside-board contribution would need preparation, relationship assessment and a calendar tested against my executive role; twenty-three days is an initial annual proposal. I would bring experience of changing a technical claim I had supported, while preserving the difference between a model that can produce an answer and evidence that the answer is suitable for a particular decision.

Current focus

Energy-planning products evaluated against information timing, hard constraints and the resolution of the decision.

Conversations I welcome

Energy-software technical leadership and a separately assessed first non-executive contribution.