Cascade, a startup building an AI platform that helps architecture, engineering, and construction firms find and win projects, has raised a $3.5 million seed round led by Andreessen Horowitz Speedrun, with participation from Ada Ventures and Snowball VC. TechCrunch reported the funding on July 22, 2026, following an initial press release on GlobeNewswire on July 21 and earlier coverage from Forbes on July 20.

Founded by Hannia Zia and Joana Ferreira and launched in 2025, Cascade operates at usecascade.ai. The platform pulls together data from U.S. state, city, district, county, and federal agency procurement portals — the websites where governments post contract opportunities — and uses it to track ongoing and upcoming construction projects. According to The Next Web, Cascade identifies multimillion-dollar projects months before any request for proposal is issued, giving firms an early lead on building their sales pipeline.

The predictive component sets Cascade apart from conventional procurement intelligence tools. The system uses historical tender data — records of past bid competitions and their outcomes — to forecast which developers are likely to win specific construction deals. As customers report back on bids they have won or lost, the AI incorporates that feedback to refine its predictions over time, creating a feedback loop that should, in principle, improve accuracy as the customer base grows.

Cascade has already secured contracts with firms responsible for building JFK Airport, LaGuardia Airport, Four Seasons hotels, and data centers. Ferreira provided firsthand details to TechCrunch, confirming the company's traction among established construction players. The startup's competitive set includes GovWin IQ and ConstructConnect, both of which operate in the construction procurement intelligence space but without Cascade's predictive win-probability modeling.

The founders applied to the a16z Speedrun program in September 2025. Speedrun, Andreessen Horowitz's accelerator for early-stage startups, led the seed round. Tech.eu and FinSMEs also covered the round on July 21.

Cascade plans to deploy the seed capital across go-to-market activities, industry events, and engineering hires. The hiring emphasis on engineers signals that the data aggregation and prediction pipeline, rather than sales infrastructure, remains the primary technical bottleneck. Scraping and normalizing procurement data across hundreds of fragmented government portals, each with its own data format and update schedule, is a data engineering problem of considerable scope.

The construction industry's procurement cycle is long, opaque, and relationship-heavy. Firms typically learn about projects through trade networks, public notice postings, and incumbent advantage. A platform that surfaces projects early and attaches win-probability estimates to specific competitors addresses a real information gap. Whether the predictive model achieves sufficient accuracy to change bidding behavior at scale is the open question, and one that the feedback-loop design explicitly defers to post-deployment iteration.

The broader context here is that AI-driven intelligence layers applied to regulated, document-heavy industries such as government procurement and construction have followed a familiar arc: the data aggregation problem is solved first, then the prediction layer follows once enough historical signal accumulates. Cascade appears to be positioning at the intersection of both, betting that early movers who capture the feedback loop will build a defensible moat through proprietary win-loss data. The risk is that incumbents like GovWin IQ, with years of accumulated procurement data, could build similar predictive features. The differentiator, if it holds, will be speed of model improvement, not data access alone.