Somewhere on a wind turbine off the coast of the Netherlands, a technician reaches for an answer. The manual runs 10,000 pages. The Wi-Fi signal died an hour ago. The pressure of a wrong decision could ground a multimillion-euro asset. Until recently, that technician had two options: guess or climb back down and search later. Twindo, a Netherlands-based startup founded by Jules Shertser, just killed both of those options.
The company has built the world’s first offline AI copilot for mobile, a small language model that runs directly on a technician’s phone via neural inference, requiring only 4GB of RAM and no internet connection. It achieved 93% accuracy and 100% satisfaction during field testing. Google invited the Twindo team to its Paris headquarters last year to study how they pulled it off.
The tech giant’s own engineers, overseeing 3.5 billion Android devices, still haven’t replicated the feat. Twindo now serves nine clients across wind energy, covering more than 21,000 turbines globally — roughly 5% of every wind turbine on the planet. The startup became the Dutch market leader in offshore wind within 18 months, recently earned ISO 27001 certification, and project revenue is surging past €1 million in 2026.
The Invisible Crisis On The Frontline
Energy infrastructure is scaling at a furious pace, but the workforce doing the hard labor is stretched dangerously thin. Technicians in wind, solar, oil and gas, and battery storage operate in some of the most isolated environments on Earth. Connectivity gaps persist. Reporting stays manual. Workflows are scattered across siloed systems. The data generated remains unstructured, buried in spreadsheets and legacy platforms inherited from the oil and gas days.
The consequences are measurable. Over 40% of asset failures trace back to human error in the field — often because workers lacked structured support at the moment of decision. An expensive technician eats into margins. Quality assurance cycles stretch across months. And the entire cloud-only AI wave crumbles the moment a signal drops on a remote offshore platform.
Jules Shertser lived this problem before he tried to fix it. A former wind turbine technician, he spent years climbing those towers, breaking fingers at 5 a.m. in the dark while the North Sea wind tore through his gear. “We were in the field, we understand the pain,” Shertser said. “We’ve had those broken fingers at 5 a.m. on the wind turbine when it’s dark and windy. We understand what is required, and it’s not always the solution that a technology architect would suggest.”
A Copilot Born From Calloused Hands
Twindo’s founding team — Shertser, CTO Frank van Luijn, and sales lead Guilherme Marinho — didn’t emerge from a Silicon Valley accelerator. They came from the job sites. Van Luijn ranks among the world’s top PHP developers and maintains one of the language’s core frameworks, powering millions of projects used by governments, airports, and companies like Slack. Shertser brought 14 years of frontline renewable energy knowledge, grease under his fingernails included.
Their first technical gamble flopped. The team spent a year building machine learning trees for predictive planning and shelved the project when it didn’t deliver meaningful gains. The lesson stung, but it reoriented everything. They stopped chasing data flows and started focusing on the humans themselves — the technicians making critical calls every day with inadequate tools. The resulting product works on two fronts.
A field copilot runs on the technician’s mobile device, powered by a custom small language model that reasons through enterprise documents locally. A technician can ask a question — voice-to-text, no buttons — and receive a contextualized answer pulled from thousands of pages of manuals and safety documents. A second, independently built neural network serves the office side, letting managers trigger agentic workflows and run analytics in seconds. Twindo coined the category “Vibe Inspections” for this dual-copilot mechanism, and the inspection process has been reduced by up to 95% in certain cases.
Standing On The Shoulders of Giants
The energy sector’s appetite for AI compute is, ironically, outstripping the planet’s ability to generate enough power. Someone has to build and maintain the new energy assets that will close that gap. And those who build and maintain them are the same overstretched, under-supported frontline workers Twindo targets.
“Technicians are standing on the shoulders of giants when they’re working,” Shertser said, referring to the colossal turbines his clients service. “And our technology is standing on the shoulders of giants as well, because we’re actually standing on top of the technicians themselves.” Twindo is now preparing to go asset-agnostic, launching across solar, oil and gas, battery storage, and, eventually, data centers in 2026.
Pilots are already running. The company has experienced zero client churn, and a seed funding round is underway to accelerate the Android version of the copilot and grow the team. Shertser envisions the offline copilot becoming the new normal for every blue-collar worker who needs a sensei in their pocket, whether the signal is strong or completely gone. An industry built by people wielding wrenches and climbing towers at dawn now has an AI that speaks their language and doesn’t need Wi-Fi to do it.
