Gilad Tal

Gilad Tal is co-founder and chief executive of DualBird, a Boston-based deep-tech company building what it calls a cloud-native data and AI infrastructure engine. The pitch is compact and slightly counterintuitive: take the kind of acceleration that normally requires custom chips, and deliver it as software that plugs into the tools companies already run. In DualBird's own framing, it is trying to rebuild the foundation of analytics and AI data infrastructure from the silicon up.

The core technology sits on rewritable hardware - field-programmable gate arrays, or FPGAs - running on Amazon's cloud instances. Instead of asking customers to rip out their stack, DualBird offers a plug-in for Apache Spark and compatibility with the Apache Iceberg table format. The claim is aggressive: up to 100 times faster performance and up to 90 percent lower cost, with no manual tuning and no change to existing workflows. Whether every workload hits those numbers is a fair question, but the ambition is clear.

Silicon as software

To understand why DualBird thinks this is possible, it helps to understand Tal's background. He trained at the Technion - Israel's storied engineering school - graduating summa cum laude in electrical engineering and physics. He then spent roughly 14 years in ASIC and software development, working on high-performance networking, compute, AI acceleration and electronic design automation. Those are the disciplines that sit at the boundary where chips meet code, and they map neatly onto the problem DualBird is chasing.

The investors backing the company have leaned into that duality. Angular Ventures, an early backer, described DualBird under the banner "silicon as software," arguing that its founders combine atomic-level silicon design knowledge with modern cloud database expertise - a pairing the firm called rare. Three of DualBird's four co-founders bridge both hardware and software, and several of the team spent time inside Amazon's own silicon and cloud operations.

A founder, not for the first time

DualBird is not Tal's first company. In 2016 he co-founded TerrainEDA, an electronic design automation startup, and served as its CEO. The company was acquired by Synopsys, one of the giants of chip-design software, and Tal continued inside Synopsys after the deal. Along the way his resume runs through names familiar to anyone who follows the semiconductor world - EZchip, known for high-performance networking silicon, and Habana Labs, the AI-accelerator company Intel bought.

That pattern matters. Building EDA tools means building the software that engineers use to design chips in the first place. Then building acceleration hardware means using those chips to make other software faster. DualBird is, in a sense, the point where both halves of his career converge: a product that treats hardware acceleration as something you deploy like any other cloud service.

The bet on data's GPU moment

The wager underneath DualBird is a historical analogy. For years, machine learning ran on CPUs until GPUs proved dramatically better suited to the math. That shift did not just speed things up - it reshaped the entire economics of AI. DualBird's argument is that data processing, the enormous and often invisible work of shuffling, joining and transforming data, is now in the position machine learning was before GPUs took over.

If that analogy holds, the prize is large. The team points to the vast build-out of data-center capacity the industry expects this decade, and positions specialized data processors as a way to do far more with far less of it. The company has said it is targeting general availability of its engine in early 2026. Skeptics will note that "specialized silicon beats CPUs" has been promised before and often stumbled on the difficulty of actually deploying it. DualBird's answer is to hide the hardware entirely behind software.

Leading through a hard chapter

Tal did not set out to be DualBird's CEO. He co-founded the company as its chief technology officer alongside Amir Gilad, who served as chief executive, and fellow founders Ehud Eliaz and Ohad Gamliel. In late 2025, shortly after the company announced its funding, Amir Gilad passed away unexpectedly. DualBird said publicly that his influence would keep shaping the company's culture and direction. Tal stepped into the chief executive role to carry the mission forward.

It is the kind of transition no founding team prepares for. What can be said from the public record is that the company kept moving - continuing to build toward general availability and holding onto the vision the founders had set together. For Tal, the shift from CTO to CEO means carrying both the technical center of gravity he has always held and the responsibility for where the company goes next.

The through-line in Tal's work is a preference for solving problems at the level where they actually live. When a workload is slow, the reflex in software is to add more machines. His instinct, shaped by years in silicon, is different: change what the machine is. DualBird is the commercial expression of that instinct - and, for now, the clearest statement of what he believes the next decade of data infrastructure should look like.