Company
We exist so that the hardest calls in a storm—the ones made at 2am with half the information—can be explained afterward, and measured against what actually happened.
DOVRANE INC. · STORM RESTORATION DECISION SUPPORTWhy this matters
Storm restoration runs on incomplete reports, conditions that change hourly, and never enough crews or materials. Every choice about what to fix next shifts the whole restoration curve behind it. Those calls get made under real pressure—and afterward, almost nobody can reconstruct why.
Storm hardening reduces how much breaks; it does not decide what to fix first when something does. That narrow problem is the entire company.
Leadership
Enterprise leadership, mission-critical platform architecture, and machine learning that respects evidence—pointed at one goal: restoration decisions operators can understand and trust.
01Founder & Chief Executive Officer
Ruchi started Dovrane on a specific conviction: the hard problem in storm response is not detecting damage, it is deciding what to repair first, and nobody was building for that decision. She owns the product thesis, the claim discipline this site is held to, and the conversations with utilities and partners. Eighteen years in enterprise technology sold into regulated, mission-critical buyers before this.
02Chief Technology Officer
Steve owns the architecture, including the parts a utility security team will push hardest on: the read-only boundary that keeps Dovrane out of operational systems, tenant isolation, and the versioned decision record that lets a recommendation be reconstructed months later with only what was known at the time. Two decades in enterprise software, distributed systems, and cloud-native platforms.
03Founding Machine Learning Engineer
Sarthak builds the perception side: the computer vision and multimodal models that read storm imagery and field reports into a ranked view of the damage. His harder problem is calibration—deciding when the model should say it does not know rather than guess, and keeping conflicting reports visible instead of averaging them into a confident answer nobody should trust.
What we work on
This is not a general-purpose AI product pointed at utilities. Each piece exists because the repair-order decision needs it.
Computer vision and multimodal models that hold on to uncertainty and conflicting reports instead of averaging them away.
Topology-aware reconstruction of an event, limited to what was genuinely knowable at each point in time.
Continuous re-planning as conditions change, with independent checks on whether a plan could actually be executed.
Versioned data, records you can follow, tenant boundaries, reproducible results, and review workflows.
Next step
Utilities, partners, investors, and engineers all reach us the same way.