The operations layer for critical facilities,
starting with data centers
Visum models data center equipment and how it is connected at each site, from the manufacturer's manuals and the site's design documents, and uses those models to train operators, write commissioning scripts and check procedures.
"Every piece of equipment in a data center has a correct way to be operated, written across manuals, drawings, commissioning reports and safety standards that share no common vocabulary. Our vision is Software Defined Operations: a working model of every unit and of every site, so that operator training, commissioning scripts and every procedure are checked against the same models, at every site, and stay current as equipment and standards change."
"Visum AI builds a working model of each piece of data center equipment, such as a coolant distribution unit or a UPS, from the manufacturer's manuals and drawings. The model knows every mode the unit can be in and refuses any step that is unsafe. For each site, Visum also builds a model of how the equipment is connected, from the site's single-line diagrams (SLDs), piping and instrumentation diagrams (P&IDs) and sequence of operations (SOO). From the equipment and site models, Visum produces operator training, Level 4 and Level 5 commissioning scripts, and site procedures that match how the equipment is installed. The same models feed all three, so what technicians are trained on and what the check enforces never drift apart."
Helping define the operations
layer, not just build on it.
We are working with standards bodies, national labs, and hyperscale operators to give AI factory operations a shared, open vocabulary, the same way Brick did for building automation a decade ago.
Built by people who have
done this before.
Nauman Rafique is the Co-founder and CEO of Visum AI. He holds a PhD in Computer Engineering from Purdue University.
Before Visum, he spent 16 years at Google, where he was the founding engineer of Google Cloud Bare Metal and grew it into a multi-team engineering organization reporting to him. The service ran globally inside third-party colocation facilities. He owned the vendor and provider relationships behind it, including a vendor team of more than 150 people that delivered the service in every global region, and was responsible for a product portfolio above $250 million. Those roles made visible the gap between equipment documentation and operational reality.
At Visum he leads the work that closes it: a model of each piece of data center equipment built from the manufacturer's manuals, and a model of how the equipment is connected at each site, built from single-line diagrams, piping and instrumentation diagrams and sequences of operations. These models produce operator training, Level 4 and Level 5 commissioning scripts, and site procedures that are checked against fixed safety rules, with no LLM deciding pass or fail.
Together with Lawrence Berkeley National Laboratory, he extended the Brick Schema ontology with a vocabulary for operational entities, procedures, interlocks and isolation points, and presented the work at OCP Global Summit 2026.
Leonardo Bachega is the Co-founder and CTO of Visum AI. He holds a PhD in Electrical and Computer Engineering from Purdue University, where his research was on modeling high dimensional signals with the Sparse Matrix Transform.
Before Visum, he spent more than 11 years building AI systems at Microsoft and Amazon. At Microsoft he was on the core teams that built Satori, the knowledge graph behind Bing. At Amazon he was a founding member of the Amazon Robotics team, where he built simulation of fulfillment center robots and delivered robotic systems into production. At AWS he led the reinforcement learning team and the model zoo teams, and led teams of PhD applied scientists building core AWS AI features. He was a founding member of AI teams that took a product from idea to eight figure revenue.
At Visum he leads the knowledge graph and the equipment and site models. He built Visum's first single line diagram parser, for the Siemens technology development partnership (it reads a site's diagram and derives the bill of materials), and designed Visum's internal format for procedures.
At OCP Global Summit 2026 he contributed to the Future Technologies Symposium paper "From Guided Technicians to Lights-Out AI Factories: A Machine-Verifiable Operations Layer", with Marco Pritoni of Lawrence Berkeley National Laboratory and Nauman Rafique.
Deep expertise across
every dimension we need.
Mission-critical operations leader with 23+ years in data centers, nuclear energy, and utility-scale power. Spent nearly nine years at Google in data center operations. Currently Director of Critical Environments at Upchurch and Co-Founder of Method Xperts.
Technology executive leading monetization at OpenAI. Founded Jams (acquired by OpenAI). Former VP of Creators and Monetization at Meta. PhD in Computer Science from Purdue University.
Associate Professor at University of Florida specializing in computer vision, robotics, and AI with focus on real-world applications.
Book a demo on
your site's equipment and design.
Book a 30-minute demo of the commissioning scripts and procedures Visum produces from models of your equipment and your site.