Heavybit Welcomes New Member: Antfly

"While AI agents have made immense strides in coding, much of the rest of the enterprise lags frustratingly behind," said Joe Ruscio, General Partner at Heavybit. "Elsewhere, information is scattered across structured and unstructured systems, with nothing indexing the right context at the right time. This is the 'retrieval gap,' and Antfly closes it with lexical, semantic, and graph retrieval in one framework, fine-grained access controls, and deployment in your own environment. James and AJ previously built data systems at billion-profile scale at Lytics, and that experience guides Antfly's design and ergonomics. Heavybit is proud to back them from the start." - Joseph Ruscio, General Partner / Heavybit

Heavybit Welcomes New Member: Antfly

Heavybit is excited to welcome the retrieval engine Antfly to our portfolio. Antfly solves one of the most important emerging problems in agentic operations today: Retrieval.

AI agents have exciting potential but in practice, they continue to be unreliable. Some studies note that 45% of all queries return errors, and that 40% of agentic projects will be canceled in the next year.

These performance challenges are so widespread that it doesn’t take formal research to recognize them. In particular, agents continue to struggle to return accurate responses to queries. The end result is simply a subpar user experience that drives away customers…but under the hood, the causes are anything but simple.

Why We’re Excited

Getting agents to consistently return excellent responses is a challenge with multiple layers that goes beyond model horsepower and harness fit. In order to build agents that return the best possible responses with any kind of consistency, operations teams also need the ability to ground their agents deeply within their own proprietary data.

To stand up retrieval for their agents, resourceful but time-constrained builders have been attempting to connect various point solutions (storage, data warehouses, models, harnesses, and other tools. Results have been mixed. Even stitching together best-in-class tools leads to severe limitations. For example, the world’s best vector database is going to be excellent at storing and managing vector data, but was never built to orchestrate retrieval across multiple third-party tools at scale.

The team has built a foundational primitive to close the agentic retrieval gap: the Antfly retrieval engine. The product finally gives teams a better alternative to jury-rigging brittle systems across multiple third-party APIs: A unified system that combines data connectivity and superior retrieval performance within a closed-circuit system, 100% owned and controlled internally from day one.

In the future, the Antfly team will continue to scale its team while enhancing its platform. Visit the Antfly website to learn more.

Meet the Founders

James McDermott, CEO and Co-Founder

James McDermott is a multi-time startup founder. He previously founded Lytics, the customer data platform for marketers, which was acquired by Contentstack in 2024, and has served in executive leadership roles at Storycode and Webtrends.

AJ Roetker, Co-Founder

AJ Roetker is a veteran engineer who has served tours of duty at Contentstack, Lytics, and Puppet.