Altas
With funding from Bill Gates, I co-founded Atlas to use machine learning to reimagine how people interact with their digital past. Unlike traditional search tools that focused on files or content, Atlas emphasized moments and context—capturing not just what you created or consumed, but when, why, how and with whom. Designed for the knowledge economy, where speed of access and contextual understanding drive value, Atlas pioneered a new category of semantic productivity tools. It was successfully exited with an acqusition by Augment.io
Challenge
Knowledge workers were drowning in fragmented digital trails scattered across devices, apps, and formats. Traditional search relied on keywords and file names, forcing users to remember where something was rather than what it meant. This created friction, slowed productivity, and increased cognitive load. The opportunity was to design a memory-like retrieval system that could resurface insights, events, and actions across a user’s entire digital ecosystem.
Solution
Atlas developed a personal semantic index, combining user activity, behavioral patterns, and content to create a searchable timeline of moments. Key innovations included:
Contextual Search – Users could reconstruct past events using images, actions, or semantic cues rather than filenames.
Mad-Lib Search UI – A natural-language style interface that allowed intuitive query building, strengthening IP defensibility.
Personalized Insights – Automated recommendations encouraged daily use, positioning Atlas as a behavioral layer on top of the OS.
Enterprise Pivot – While initially B2C, early traction among knowledge workers and executive assistants revealed enterprise potential in collaboration and knowledge management.
Results
Validated Product-Market Fit through alpha and beta adoption with knowledge workers and enterprise teams.
Investor Confidence – Direct engagement with Bill Gates, Nathan Myhrvold, and other backers led to continued support and funding.
Market Expansion – Successfully pivoted from consumer to enterprise, focusing on high-value knowledge industries.
Proprietary Innovation – Established defensible IP through unique visualizations and semantic UI patterns.
Monetization Strategy – Validated pricing models and revenue streams for enterprise deployment.
My Role
As Co-Founder, I was responsible for shaping both the product and the company strategy:
Defined the product vision, roadmap, and requirements.
Directed iterative design and user research to develop the semantic search interface, visualizations, and index tools.
Established a biweekly user research cadence, identifying trust metrics that drove adoption.
Built the design language and comprehensive design system.
Co-led fundraising for subsequent rounds.
Drove GTM strategy, collateral, and sales materials.
Led the pivot from B2C to B2B, validating enterprise market fit.
Capturing the moment
Given Atlas captured running index of all content on your device and actions you took, then stored that information on a timeline. Traditional searches find an item—a message or webpage, they search for needles in a haystack; Atlas was a time machine.
Altas allowed you to easily move from that one message to recreating the moment you first read it—opening that excel file, your browser tabs, related emails, even the music you were listening to.
Altas integrated with OS and internet search tools as well. Since Altas’ index was driven by what the user had done or seen, rather than a general public index driven by popularity or promotional algorithms, Altas’ results were truly personalized and they could remind you of things you may have already read, done, or seen related to that Google or Apple search.
Visual Search
Rather than relying on a list format of results, Atlas employed a visual search, displaying images from your activities grouped together by time, topic, people, locations, etc. Atlas created semantic connections based on behavioral patterns, activity loops, topics, locations, and routines to capture the moment rather individual items.
By indexing everything you were working at any moment, Atlas captured interactions between applications and websites (i.e., copy/paste, flipping between windows, etc.). Opening a link from an email or message Atlas would track the sender, the other recipients, your reply, and what you did next. Allowing Atlas to return you to that moment, what you were doing, and who you were interacting with, augmenting your memory and recommending next steps.
Mad-Lib Search Interface
Another design innovation for Atlas was the use of a mad-lib style search interface. The user could type or select from a set of words or short phrases to focus or expand their search. This allowed the user to search independent of keywords. Once they had returned to the moment in time they were looking for, they could then share the moment with friends and colleagues, or simply pick-up where they left off.
The Mad-Lib style interface allowed users to have discrete control over their search. The Mad-Lib set could also be tailored to spend content types, locations, or different contexts.
Reliving a moment to plan the next
Traditional search engines index single items—a message, a webpage, a file; allowing you search for needles in a haystack; Atlas was a time machine.
Altas allowed you to easily move back in time starting with a key memory—a message, a person, that website you found, and from there recreate the moment you first read it—opening that excel file, your browser tabs, related emails, even the music you were listening to.
Since Altas’ index was driven by what the user had done or seen, rather than a general public index driven by popularity or promotional algorithms, Altas’ results were truly personalized. Unlike other search services, Atlas would know that you not only looked at those ski boots, but that your purchased them, while chatting with your friends about when and where you could skiing, and even what you had planned for dinner the first night. Then it could build out recommendations, pulling together information from the internet, your past likes, etc.