Built for the highest-stakes projects on the planet

My career began with a question: how do you unlock the value of critical data without losing control of it?

As an engineering mathematician in corporate R&D, I began investigating this question in 2001 while spearheading work in machine learning and data privacy. My published work (scientific papers & patents) made a global impact, however, I didn't want those ideas to stay confined to research; I wanted to build systems that solved real-world problems and, in 2016, I transitioned into commercial data science development.

In 2019, my path naturally aligned when I discovered the art of large capital project management. For the next 5½ years, I led AI and Engineering, at Nodes & Links, where my team applied AI and mathematical modeling to mega-projects. We examined how delays—even those outside the critical path—could ripple through a billion-dollar schedule.

That experience revealed two things.

Firstly, project data contains enormous untapped value, and the organizations that need that value most are often the least able to risk exposing it. I sat across the table with PMO Directors and Project Controllers who needed AI-grade answers, only to watch them hesitate; handing over that data was a risk they simply couldn't take. Confidentiality clauses, contractor liability, sovereign data rules—the sensitivity of capital project data isn't a compliance footnote. It's a material business risk.

Secondly, tedious workflows are consuming the time and attention of PMO teams. Project Managers and Controllers need faster analysis, automated reporting, earlier visibility of risks and delays, and a connected view of schedule, progress, cost, risk and project health. They also need to coordinate actions across teams without stitching together separate tools.

Crucially, bolting an AI assistant onto fragmented project controls does not fix the fragmentation underneath. AI must be embedded within the control environment—working across connected data to accelerate analysis and reporting, surface important relationships, and support managerial judgement without replacing it.

Fixing this fragmentation safely requires a complete rethink of how software is built. I discovered that the classic SaaS architecture is not always the best technology for security and performance engineering: ship the data out, wait for a server somewhere to queue it, store it, process it, and ship an answer back. While cloud software transformed enterprise collaboration, it has not been designed for every type of data or every operational context.

Recognizing this gap is how we forged our vision at Privalyx—and I'll say it plainly: for critical and commercially sensitive work, cloud-only software (SaaS) should no longer be the default. It will give way to a local-first architecture in which sensitive data and critical computation remain on hardware the customer controls. When cloud services are needed for collaboration, backup, additional computing capacity or external LLMs, privacy must remain enforced by design.

That's the future we are building towards: a world where the most sensitive, highest-stakes projects on the planet run on AI that's as private as it is powerful. Where “privacy-first” doesn't mean “slower” or “less capable,” and where project teams never again have to choose between getting the answer fast and keeping their data safe.

Privalyx exists because we refused to accept that trade-off.

— Dr. Georgios Kalogridis, Founder, Privalyx