We're building the brain for thermal infrastructure.
HeatVX was founded in 2025 to productize what Google DeepMind proved internally — that AI can cut cooling energy by 40% — and make it available to every data center, hyperscaler, and industrial operator on earth.
From research lab to control loop.
In 2016, Google DeepMind quietly cut cooling energy at one of Google's data centers by 40% using reinforcement learning. The result was published, celebrated, and then locked away inside Alphabet.
For nearly a decade, the rest of the industry had to choose between two bad options: buy expensive proprietary thermal cameras with no analytics, or stitch together SCADA systems that take 6 months to deploy and don't learn.
Then AI racks happened. GPU densities shot from 15kW to 100kW+ with 500kW on the roadmap. 83% of data center experts said the cooling supply chain couldn't keep up. The crisis was no longer theoretical.
HeatVX exists because the world can't afford to wait another decade. We're the pure-software AI brain that turns any thermal sensor into predictive intelligence — no hardware swap required, deployed in days, paid back in months.
Eliminate wasted heat.
Empower every operator to optimize thermal performance through real-time AI — reducing energy costs 30%+ while preventing catastrophic failures.
Become the global standard.
Be the AI brain behind every thermal sensor on earth — the operating system for industrial heat, from data centers to refineries to fabs.
What we believe.
Builder-first
Engineers run the company. Every leader can read a stack trace at 2am.
Customer-aligned
We win when our customers cut energy bills. Pricing scales with their savings.
Ship fast, ship safe
Two-week release cadence. Hard safety bounds on every closed-loop action.
Intelligence amplified
AI augments operators — not replaces them. Humans always have the kill switch.
Builders, operators, researchers.
Ex-Google DeepMind AI4Energy. 10y building AI for industrial systems.
Ex-NVIDIA DGX cooling lead. Built thermal control loops at hyperscale.
Ex-Stripe infra. Took payments from 100 to 100M req/s. Doing it again for thermal.
Ex-Linear. Believes great B2B tools are indistinguishable from great consumer ones.
PhD MIT. 40+ papers on physics-informed neural networks.
Ex-Vertiv, Schneider. 15y selling mission-critical infra to data center operators.
Open-source maintainer. Built thermal-py (5k stars). Loves a good CLI.
Ex-Stripe Press. Believes infra deserves the same craft as consumer apps.
