Guava Raises $12.5M Seed for Developer Voice AI

Guava puts a forward-deployed engineer on its enterprise price sheet. That is a revealing line for a company selling a voice API: the purchase includes someone who can help carry the customer's business rules into a working conversation.

The Los Angeles developer platform announced a $12.5M Seed round on November 3, 2026, led by Acrew Capital. Flourish Ventures, WillowTree Ventures, Clocktower Ventures and returning pre-seed investor Commerce Ventures participated. Guava says the capital will support model and product development and help build its go-to-market team for enterprises and business process outsourcers.

The commercial proposition reaches beyond a voice that sounds convincing. Guava owns the speech models and runtime; its customers retain control of business logic through their own code and systems. The enterprise integration work sits where those responsibilities meet, which makes the people delivering the API part of the funding story.

What Guava's Seed round is financing

Guava develops and serves speech recognition, language models and speech synthesis as an integrated system. Its developer tools let customers connect their knowledge and business rules while Guava manages the voice interaction. An engineering team can build around Python and TypeScript rather than take responsibility for assembling every part of the speech pipeline.

That choice creates a substantial assignment for the supplier. Owning the models means Guava can work across the system when a conversation needs improvement, but it also keeps the model-development work inside the company. The funding supports both that technical work and the commercial capacity needed to bring it into more organizations.

Guava's release identifies edge-device applications, on-premises installations and full-duplex interaction as development directions. Full duplex would allow an agent to listen and speak at the same time. These are plans described with the financing, rather than a claim that every deployment already includes those capabilities.

The company calls the financing a Seed round and identifies Commerce Ventures as an existing pre-seed investor. The announcement does not disclose the earlier round's amount, Guava's cumulative funding or a valuation. The $12.5M figure therefore describes this announced transaction; adding an assumed earlier cheque would manufacture a total the record does not supply.

Four founders and a commercial handoff

Guava's leadership team combines Omri Dahan, co-founder and CEO; Anthony Scodary and Nico Benitez, co-founders and co-heads of engineering; and Mike Piechna, co-founder and CRO. The engineering titles matter: Guava lists two co-heads, rather than naming a CTO. Piechna's revenue responsibility also belongs in a story about building a commercial organization.

Dahan previously spent more than a decade as Marqeta's CRO, helping grow the payments company from its first revenue through its public-market filing, according to Guava's release. Before becoming a Guava operator, Dahan had invested in Gridspace. Scodary and Benitez were technical co-founders of that voice research business, giving the new commercial assignment a much longer technical history behind it.

That history needs careful accounting. Guava describes its platform as built on more than a decade of Gridspace model development, including training on more than 10B minutes of live conversations in regulated call centers. Those figures describe the research lineage presented by the company. They cannot establish Guava's incorporation date or be quietly converted into a current Guava customer count.

For an enterprise buyer, the combination gives the vendor conversation two owners: the people improving the models and the people responsible for getting a deployment into service. Building the sales team expands the number of organizations Guava can approach. Keeping engineering close to that work determines how much of the original research a customer can actually use.

Business rules remain the customer's job

The useful distinction in Guava's product architecture is where control lives. The company describes business rules written in code, alongside existing customer systems such as a CRM or electronic health record. Guava supplies the conversational voice layer around those rules, allowing the institution's own software to remain the source of truth.

This makes review and maintenance part of the product's appeal. A developer changing an agent's permitted behavior needs to understand which rule changed and how to test it. Keeping that work near an existing codebase gives the team a familiar place to manage changes, even though the conversation reaching the caller is generated by AI.

A buyer still has to connect the rules to a particular operation. Guava's enterprise offering includes a named account manager, a forward-deployed engineer, dedicated support and infrastructure options. Those are specific commitments around the deployment, and they help explain why an API company needs commercial and implementation capacity as it grows.

Guava is consequently asking customers to evaluate two related purchases: the voice technology and the working relationship around it. A developer can begin with the self-service offering. An institution putting that voice inside a regulated workflow has a longer conversation about integration, support and the responsibilities each side is taking on.

Daytona's early activity and the work ahead

Guava says Daytona, its first public model, launched in September and already supports more than a dozen enterprise customers, handling tens of thousands of calls each day. These are company-reported figures in the funding announcement. They give the round a production setting without supplying independently audited revenue, retention or customer-outcome data.

The company's current hiring page makes the organizational buildout tangible. It lists roles across machine learning, security, quality assurance, enterprise sales, forward-deployed engineering and product-led growth, alongside operating and governance functions. The job mix follows the assignment described in the financing: develop the system while building the organization that sells and supports it.

Guava is hiring for both routes into that relationship. Its product-led growth role is charged with getting a developer from discovery to a call with as little sales involvement as possible, while enterprise sellers work across technical and business buyers. Dahan and Piechna are building a company that has to make room for both: the engineer starting alone and the institution arriving with a roomful of people.