About — Capital Project AI

Why we built Capital Project AI

Capital committees make some of the highest-stakes decisions in the economy — which projects to fund, how much contingency to carry, when to commit to long-lead vendors, when to pull back. The math that should drive those decisions is well understood: probabilistic forecasting, portfolio optimization, reference-class calibration. In practice, almost none of it makes it into the sanction case. The committee sees a deterministic plan, a risk register, and whoever made the most persuasive argument in the pre-read meeting.

We started Capital Project AI because the founding team had seen the gap from both sides. An ex-Shell capital owner who had watched $800M of megaprojects miss their sanctioned outcomes in predictable ways. A Penn State professor whose peer-reviewed work on capital allocation optimization showed consistent 2.7x return improvement over traditional methods but sat unused outside academia. The platform is the operationalization of that research, packaged around the way capital committees actually work.

$800M+

of delivered capital project experience in the founding team

2.7x

return improvement vs. traditional methods in peer-reviewed studies of 95 industrial projects

25 years

of academic research on capital allocation optimization informing the platform

The founding team

Dr. Abhishek Kar — Founder & CEO

PhD in Chemical Engineering. Eleven years at Shell leading capital projects worth $800M+, including the PAO market entry. Direct operational experience on the projects where capital decisions get made and where they go wrong. Combines deep engineering domain expertise with quantitative finance and AI/ML technical depth — the rare combination this category demands.

Dr. Darrell Velegol — Founding Advisor

Distinguished Professor of Chemical Engineering at Penn State. The world's leading researcher on capital allocation optimization for industrial portfolios. His peer-reviewed work across 95 industrial projects — published in Industrial & Engineering Chemistry Research (2025) — is the academic foundation that the Optimize engine operationalizes.

Why this combination matters

Capital project software built by software people tends to optimize the wrong thing — activity completion, report generation, dashboard density. Capital project software built by operators tends to stop at scheduling. What this category actually needs is the combination: somebody who has lived the decisions, and somebody whose research has identified the math that makes those decisions measurably better. Capital Project AI is built on that intersection.

What we believe

Capital allocation is the largest unmeasured lever in capital-intensive industries. Cost programs squeeze a few percent from each project. Allocation changes which projects get funded. The second lever is 5-10x larger, and it's where we focus.
Probabilistic beats deterministic every time. A plan that says "$200M and 24 months" is lying. A plan that says "P50 $215M, P80 $260M, with these three drivers" is honest. Committees that sanction on the first and manage against the second quietly lose money for a decade.
The math exists. What's missing is operationalization. Monte Carlo, reference-class forecasting, fractional Kelly, constrained optimization — none of this is new. The gap is getting it from research papers into the meeting where the decision is made. That's what the platform is.
Built for the committee, not the scheduler. The outputs match how decisions actually get made: priority lists, trade-off tables, sensitivity walk-throughs. Not activity-code variance reports.

Where we are

Capital Project AI is headquartered in Houston, Texas — close to the capital-intensive industries we serve. We work with operators, EPCs, and infrastructure investors across North America, Europe, and Asia-Pacific. Remote by default for the technical team, with domain engagement in-person at customer sites.

We're backed by investors who understand the category — capital-intensive industries, quantitative finance, AI for regulated decisions — and we're deliberately growing slowly enough that every customer engagement goes through the founding team.