Todd
Mozer

Running a 30-Year Overnight Success

In 1994, Todd Mozer started a company that would one day put its technology inside over three billion consumer devices - and keep its own name off every single one. That was the business model. That was the genius. While the world argued about which AI company was going to win the cloud, Mozer was quietly winning the device.

Sensory, Inc. runs on-device. Its code lives in your earbuds, your car, your smart doorbell, your TV remote. When you say "Hey Siri" or "OK Google," the first thing that happens - before any cloud, any server, any data center - is a tiny Sensory-built model wakes up on the chip and decides whether you actually said that. Privacy by architecture. Intelligence without a network. Speed without latency. Three things the industry spent decades promising and mostly failed to deliver.

Mozer built that from a company that raised $400,000. Once. In 2001. And never needed more.

A Family Business Built on Physics and Code

The Mozer story starts in a physics lab. Todd's father, Forrest Mozer, is a pioneering space plasma physicist who built instruments to measure electric fields in the magnetosphere. He also had a knack for commerce. In 1983, father and son - along with Joe Costello - co-founded ESS Technology, then called Electronic Speech Systems. It was an early chip-level bet on voice that most people thought was eccentric at best.

Todd went off to Stanford for an MBA in 1986, then spent years sharpening his strategic instincts at consulting firm Strategic Decision Group, where he worked with multi-billion dollar organizations. He ran international sales at Opcode Systems, a music software company where voice-driven interfaces were ahead of their time. By 1991, he was VP of Business Development. By 1994, he and his father and his brother Mike had co-founded Sensory.

The first chips had 2K of SRAM and 64K of ROM. Neural networks on hardware that would barely run a screensaver today. People thought it was weird. "They'd ask, 'Why are you using neural nets for speech recognition?'" Mozer recalls. "And of course, that's the big thing now."

The GirlTech Moment (and the Mars Bonus)

Sensory's first big consumer hit was not glamorous. It was a pink diary with a voice lock. The GirlTech Password Journal let pre-teen girls protect their secrets with their voice. Sensory's speaker verification sat at the core. At peak, one in three American pre-teen girls owned or played with one. That is not a metaphor. That is a documented market penetration figure for a company of 49 people.

Around the same time - 1999 - the Mozer family did something stranger still. Sensory donated a RSC-164 speech recognition chip to NASA. It flew to Mars aboard the Mars Polar Lander. A chip with neural net speech recognition, traveling 309 million miles on a government rocket, while its inventors were still working on voice-controlled toys.

Forrest Mozer, for his part, also helped build the microphone that went on that same lander. The family sent two components to Mars. One for sound input. One for audio intelligence. The combination was decades ahead of anything that would become mainstream.

Building the Invisible Infrastructure

The dirty secret of voice AI in consumer electronics is that Sensory's name almost never appears on packaging. Mozer built a B2B licensing machine that powers product lines from Amazon, Google, Samsung, LG, Sony, Toshiba, Huawei, Nokia, and hundreds of others. When Bluetooth headsets had a return rate of 30-35% in the early 2000s - because the voice recognition was so bad users just gave up - Sensory's BlueGenie technology came in and fixed the problem. Same pattern, different device, different decade.

The TrulyHandsfree wake word engine became the architectural foundation for how wake words work in modern devices. The TrulySecure biometric layer added voice and face verification. TrulyNatural brought large-vocabulary speech recognition that could handle open-ended dictation. The branding was always internal. The product was always someone else's.

There is a discipline in that choice. Mozer understood early that the licensing model - where Sensory's value is measured in device counts rather than brand recognition - scales without the marketing budget. You don't need to convince consumers. You need to convince one Samsung VP, and then ship in 200 million phones.

The Privacy Bet That Took 25 Years to Pay Off

When Sensory was founded, on-device AI was not a privacy play. It was a capability play - the cloud barely existed in a useful form, so if you wanted voice recognition in a toy, it had to run on the device. The privacy advantages were incidental.

What changed is that the cloud arrived, and consumers discovered what that meant for their data. Mozer had spent two decades building AI that stayed on the device. GDPR arrived in 2018 and CCPA followed. Suddenly, "we don't send your data anywhere" was a regulatory differentiator, not just a technical footnote. Sensory had the architecture already in place. The market caught up to where the company started.

"There's a movement towards privacy, which brings things more towards on-device," Mozer noted - with the quiet confidence of someone who picked a position in 1994 and watched the world eventually agree.

Smart Wakewords: The Next Thirty Years

In December 2025, Sensory launched Smart Wakewords. The technology uses a multi-layer architecture - on-chip detection, on-device NLU validation, and cloud LLM confirmation for critical tasks - to let devices understand context, intent, and identity. A device running Smart Wakewords can distinguish between "Hey Device, what's the weather?" and someone in the background having an unrelated conversation. It can adapt to accents, noise levels, and specific users. Follow-up questions don't require re-triggering the wake word.

The launch timeline is revealing: available for licensing in January 2026, already optimized for Qualcomm's Snapdragon Wear Elite in March 2026, and already on the conference circuit by May 2026. Mozer is moving Sensory's 30 years of embedded AI expertise directly into the large language model era - solving the power and latency problems that make always-on conversational AI currently impossible on most devices.

At CONNECTIONS 2026, Mozer joined panelists from Amazon Web Services, ADT, and Arity to address the smart home experience gap. The company that spent three decades making AI invisible is now at the center of the conversation about what AI should become next.

What 30 Years Without a Pivot Looks Like

The technology industry celebrates pivots. It funds discontinuity. Mozer built something stranger: a company with a single thesis - AI belongs on the device, not the cloud - that has been correct for three decades and correct for different reasons in each decade. In the 1990s, it was correct because there was no cloud. In the 2000s, it was correct because power constraints were severe. In the 2010s, it was correct because privacy became a concern. In the 2020s, it is correct because edge AI chips have finally caught up to what Sensory's architecture always assumed they'd become.

Mozer's $400K in outside funding and $20M in annual revenue tells the story of a company that grew by being indispensable, not by being famous. The 49 people at Sensory have built the layer below the layer that consumers see. When you thank your voice assistant, you are indirectly thanking a team in Santa Clara that figured out how to run neural networks on a chip the size of a fingernail, while that chip is drawing less power than a blinking LED.

That is not a footnote. That is a foundation.