Hannia Zia and Joana Ferreira have raised $3.5 million for Cascade, an AI platform designed to help architecture, engineering and construction firms identify projects before a formal request for proposals appears.
The financing, announced Tuesday, came from Andreessen Horowitz Speedrun, Ada Ventures, Blitzscaling Ventures, Indico Capital, shuckerVC, G2C Ventures and Snowball VC. Cascade did not disclose its valuation, revenue, customer count or the date the financing closed.
Zia and Ferreira met at UnlikelyAI, where Zia was vice president of product and Ferreira led the AI platform. They worked on a knowledge graph and agents that could navigate large bodies of information, experience they are now applying to the fragmented records that precede construction projects.
Their route into the industry was personal. Ferreira grew up in a Portuguese family of carpenters. Zia's father attempted to start a construction business that failed. Conversations with construction finance executives later led Zia to Munoz Engineering, Cascade's first customer, according to the funding announcement.
Finding the project before the project exists
Cascade says it continuously analyzes bond filings, permits, capital plans, property transactions, earnings transcripts, government budgets and meeting minutes. The software connects those events to prospective construction work, scores opportunities against a customer's capabilities and looks for potential introductions through relationships stored in Outlook and other systems.
The basic thesis is that a formal RFP arrives late in the commercial process. By then, an owner may have secured financing, assembled advisers, selected likely partners or shaped a project around firms already in the conversation. Cascade wants to give smaller firms the same early visibility that larger competitors derive from local relationships and institutional memory.
Cascade says its customers have surfaced more than $10 billion in potential projects through the platform. That figure measures identified opportunity value, rather than revenue won by customers or revenue earned by Cascade. The clearest conversion example in the announcement comes from Smallwood principal Tim Johannesson, who said the architecture firm won a $6 million project it found through Cascade. That account remains a customer testimonial rather than independently reported contract data.
Cascade's public product page also shows how much of the broader vision remains to be delivered. The current platform description lists government project discovery, proposal writing and project specification analysis as live. Pre-bid intelligence, private-market signals, bid leveling and a feature that matches firms into bidding teams are labeled "soon."
The distinction matters because pre-RFP prediction is the sharpest part of Cascade's pitch. The product available today appears more developed around finding and qualifying published government opportunities, then helping firms prepare responses. The $3.5 million gives Zia and Ferreira room to build the earlier detection layer into a generally available product rather than a set of early deployments and planned features.
The data becomes the product
Cascade was founded in 2025, according to its a16z Speedrun profile. The profile says Zia previously helped scale Google Pay India from zero to 100 million users, while Ferreira worked on machine learning for Google Search and was an early engineer at trading app Freetrade. Ferreira also worked at Amazon before leading the AI platform at UnlikelyAI.
Those credentials explain part of the investor interest. Cascade's larger bet is that the records generated before, during and after each pursuit can become a proprietary data set. A customer's wins, losses, relationships and project history can improve how Cascade ranks the next opportunity. More participating firms could also give Cascade a better view of which architects, engineers and contractors make credible bidding partners.
Ada Ventures' investment memo says Cascade charges a subscription and a success fee when a customer wins work. Ada also claims early customers have closed bids within one to three weeks, compared with what it describes as a six-month industry norm, and that Cascade has recorded no churn. Cascade has not published the underlying customer cohort or the period covered by those figures.
The outcome-based fee gives Cascade a direct interest in the quality of its recommendations. It also raises the burden of attribution: on a relationship-driven construction pursuit, Cascade will need to show when its signal or introduction materially contributed to a win.
A contested pre-RFP market
Cascade is entering a category with established databases and several AI-native competitors. Citylitics analyzes public infrastructure planning and funding activity before procurement. Scaffold AI monitors agency meetings, funding announcements and capital improvement plans. Mercator.ai uses property transactions, rezoning records and permits to detect private construction projects early.
Kantiv approaches the pursuit from inside the AEC firm, organizing proposal history, project experience and client relationships so that institutional knowledge can be reused. Cascade is trying to connect both sides: external signals showing where work is forming and internal data showing whether a particular firm can win it.
That combination gives Cascade a credible wedge, provided the prediction layer produces opportunities customers cannot get from existing bid portals or competing intelligence services. The round buys the founders time to prove that advantage and expand the customer network feeding it.
Zia and Ferreira have chosen a measurable problem. A recommendation either reaches a customer early enough to shape a pursuit or it does not. A project either fits the customer's record or wastes a proposal team's time. Cascade's progress will be visible in contracts won, not the aggregate value of opportunities passing through its software.