A roof is a physical asset with a predictable failure curve. The cost of missing that curve runs into the billions, spread across insurance claims, emergency repairs, and lost asset value. RAKE ML, a New York-based startup founded last year, is betting its causal AI model can see the curve before it breaks.
The company calls itself the first vertically-integrated AI platform for the built environment [rakeml.com, retrieved 2024]. Its wedge is narrow: predicting roof degradation by simulating millions of physically-grounded failure scenarios, then translating those signals into actionable risk scores [rakeml.com, retrieved 2024]. For buyers in property insurance and large-scale asset management, the promise is a shift from reactive inspection to computational foresight.
The Wedge: Physics Over Pixels
RAKE ML's differentiation rests on a claim of causal understanding. Rather than simply identifying visible damage from satellite imagery, the platform asserts it models the fundamental physics of why materials fail [Perplexity Sonar Pro Brief, retrieved 2024]. This means simulating the impact of specific weather patterns, material fatigue, and construction flaws over time. The output is not just a snapshot of current condition, but a forecast of degradation timing and severity.
The platform targets three core decision points in the property lifecycle, translating complex signals into clear next steps [rakeml.com, retrieved 2024]:
- Pre-acquisition due diligence. Identifying roof and envelope risk before a deal closes to support pricing and negotiation.
- Insurance underwriting. Providing objective condition evidence to bind or renew policies.
- Capital planning. Informing the timing and scope of repair or replacement budgets for asset managers.
The Team and the Early Signal
The founding team brings a blend of operational discipline and technical focus. CEO Rafael Ortega is a former U.S. Naval Aviator and University of Chicago Booth MBA [LinkedIn, retrieved 2026]. CTO Kevin Kabaria, based in New York, leads the technical build [LinkedIn, retrieved 2026]. The company is operating in a classic pre-seeed mode: lean, focused on product development, and without publicly disclosed funding rounds or named enterprise customers.
A signal of early commercial motion appeared in a May 2025 LinkedIn post, where the company noted it was "expanding [its] commercial pipeline with Risk Signal Solutions" [LinkedIn, May 2025]. The post did not name specific clients, but it indicates a focus on business development. The absence of a public fundraising announcement suggests the founders may be bootstrapping or operating with a small friends-and-family round while they prove initial product-market fit.
The Counterfactual: Proving the Model
The ambition is clear, but the path is lined with execution risks common to deep-tech applied AI. The strongest counter-bet is that the real-world performance of a physics-based model cannot outpace simpler, pattern-recognition AI trained on vast historical claims data. Insurers and asset managers are notoriously conservative buyers; convincing them to trust a novel algorithmic assessment over a human inspector's report is a high bar.
The risks break down into three parallel challenges:
- Model validation. The platform's predictive accuracy must be demonstrably superior to existing methods in live, paid pilots. No public case studies yet confirm this.
- Data integration. Success requires clean ingestion of disparate geospatial, weather, and property data feeds, a non-trivial engineering hurdle.
- Sales motion. The company must build a commercial team capable of navigating long enterprise sales cycles in regulated industries like insurance.
For Ortega and Kabaria, the next twelve months will be about converting technical promise into commercial proof. That likely means securing a first institutional funding round to scale engineering and go-to-market efforts, and landing a flagship customer in insurance or commercial real estate to serve as a referenceable case.
RAKE ML's bet is that the roof, and every physical asset beneath it, is a computational problem waiting to be solved. The question for 2025 is which insurer or asset manager will be first to write a check based on that prediction.
Sources
- [rakeml.com, retrieved 2024] RAKE ML, Predict & Prevent Asset Failure Before It Happens | https://www.rakeml.com
- [Perplexity Sonar Pro Brief, retrieved 2024] RAKE ML company brief
- [LinkedIn, May 2025] RAKE ML's Post - startups #founders #ai #builtenvironment | https://www.linkedin.com/posts/rake-ml_startups-founders-ai-activity-7413987709838311426-CXjB
- [LinkedIn, retrieved 2026] Rafael Ortega - RAKE ML | LinkedIn | https://www.linkedin.com/in/rafael-ortega-736b1a222/
- [LinkedIn, retrieved 2026] Kevin Kabaria - Co-Founder & CTO at RAKE ML | https://www.linkedin.com/in/kevin-kabaria-4729019b/