Who builds HYDRA OS

Polestar Technology · Diyarbakır, Türkiye · Incorporated 2023 · Founded by Mert Satıcı

Polestar Technology is an electrolyzer intelligence company based in Diyarbakır, Türkiye. It builds HYDRA OS, an AI operating system for PEM, AWE and SOEC water electrolyzers that runs as a digital twin alongside the physical stack, forecasts degradation-driven failures up to seven days ahead, and targets a 15% reduction in the levelized cost of hydrogen.

Why this company exists

Green hydrogen does not have an energy problem, and it does not really have a technology problem. It has an evidence problem. The electrolyzer is the core asset of a project that costs hundreds of millions of dollars, and in practice it runs as a black box — dashboards and alarms, but no continuously validated, auditable record of electrochemical behaviour that survives a lender's due diligence.

That gap is what Polestar Technology was built to close. Not another monitoring dashboard, but a physics engine that can say what a stack is actually doing, with the working shown.

From July 2020 to a company

The engineering work started in July 2020. Polestar Technology was incorporated in 2023, once it was clear the thing being built was a product and not a research project.

July 2020

The question

Conversations with electrolyzer operators and developers across South Korea, China, Europe and the US. Not pitching — listening. The same answer kept coming back in different languages: nobody could tell a lender's technical advisor what the stack was actually doing.

2021

First physics models

Initial Butler-Volmer efficiency models. They returned roughly 93% HHV efficiency — impressive numbers that were also completely wrong. Real-world NREL benchmarks were considerably more sobering.

2022

Calibration against NREL benchmarks

The physics engine was rebuilt around NREL/TP-5700-81257 benchmarks, incorporating ohmic resistance, exchange current density calibration, bubble overpotential and contact resistance. Sub-5% voltage error on four of five benchmarks.

2023

Multi-agent architecture

The agent layer was added — not because "AI" was fashionable, but because monitoring, fault diagnosis and operational optimization are structurally different problems that need to run independently and coordinate, rather than share one monolithic model.

2024

The bankability reframe

Understanding how project finance actually works changed the product. The output was not just operational intelligence; it was the evidence layer lenders need to price technology risk accurately.

2025–2026

Lab validation and first pilots

Testing on a molten carbonate electrolyzer in a controlled lab environment, 54 of 54 validation tests passed after final calibration fixes, and pilot discussions opened with industrial electrolyzer operators. Presenting at Hydrogen Technology Expo, Düsseldorf, September 2026.

How we talk about numbers

The performance figures published across this site — 15%+ LCOH reduction, 30–40% stack lifetime extension, seven-day failure prediction lead time — are engineering targets derived from physics models, not measured results from a deployed customer fleet. Where a number came from a benchmark or a lab test, that is stated. Field-validated figures will be published when the Founding Pilot Cohort completes.

The same principle drives the free engineering tools: every default coefficient is stated with its source, every assumption the software makes on your behalf is listed next to the number it changed, and each tool carries a section on what it deliberately does not do. An industry that cannot agree on units will not agree on costs either.

Contact