Raiza Martin

Raiza Martin is the co-founder of Huxe and the former product leader behind Google’s NotebookLM. She builds AI products around information people already care about: first by helping users understand their own documents, then by turning news, newsletters, and personal context into customized audio.

Martin began her product career at startups before joining Google. Having previously left college, she returned to school while working full-time and earned a bachelor’s degree from Harvard. Managing coursework alongside her job gave her a concrete problem to solve: she wanted software that could absorb documents, connect ideas, and answer questions about the material.

At Google Labs, she helped develop Project Tailwind, introduced at Google I/O in 2023, into NotebookLM. Her announcement with Steven Johnson described a notebook grounded in users’ selected sources, with summaries, questions and answers, and citations. That emphasis on source-grounded AI made responses more relevant and gave users a way to verify them.

Summarization exposed the gap between an impressive model and a dependable product. Early context limits made requests about specific documents unreliable; automatic summaries ultimately provided immediate value while teaching users how an unfamiliar document-centered interface worked.

Martin left Google in December 2024 with fellow NotebookLM contributors Jason Spielman and Stephen Hughes. They founded Huxe, tested several product directions, and concentrated on audio after finding that users consistently preferred it. The company publicly launched its application in September 2025 after raising $4.6 million.

Huxe produces personalized briefings and interactive audio using contextual information such as news, newsletters, calendars, and email. Martin envisions proactive listening experiences that anticipate what someone needs without requiring another screen or repeated prompts.

Her defining product principles

• Solve one recognizable problem. Martin criticizes AI demo disease: products that showcase model capabilities without delivering a clear, repeatable user outcome. An early Huxe prototype combined chat, podcasts, news, images, and video; watching users return to one activity convinced her that product restraint creates a stronger competitive position.

• Earn trust before introducing novelty. Reliable summaries, understandable failures, visible sources, and respect for user data and agency matter more than impressive first demonstrations. NotebookLM’s generated audio became compelling because listeners recognized their own materials and remained active participants.

• Design around personal context. Both NotebookLM and Huxe start with information meaningful to an individual. Martin’s vision for personalized content extends that approach from uploaded documents to daily routines, information backlogs, and audio that adapts to its listener.

As AI tools blur distinctions between product management, engineering, and design, Martin places increasing weight on judgment: defining the user’s actual need, choosing a focused outcome, and making the product dependable enough to earn a place in everyday life.