Guaranteeing privacy for data AND algorithms. Trust us, but believe the math!

Safe Paths is a non-profit and open source organization building technology for citizens and healthcare authorities around the globe to cooperate in the the battle against COVID-19. We created a system allowing incredibly detailed contact tracing without creating tools for a surveillance-state. In conjunction with Ramesh Raskar at MIT I designed and built the first version of the technology behind Private Kit and Safe Places, then launched COVID Safe Paths with an all-star global open source team. I also had the privilege of helping to launch the initial implementations around the world with the backing of my colleagues at TripleBlind.

Mycroft AI is dedicated to the lofty goal of building the first Open Source AI. We believe the technologies needed to achieve this goal are complex enough and important enough that they should be developed collaboratively. And we believe making this technology available to all allows many innovative companies to emerge and build systems which would be impossible to create on their own.Initially we are focusing on several practical technologies: * Wakeword spotting * Speech recognition * Text to Speech * Multi-lingual natural language intent parsing * Multi-platform voice interaction frameworkWe have achieved this with several machine learning-based solutions we created -- data acquisition, collaborative (crowd-sourced) data tagging, automated training and production deployment of learned models. These combine to empower even small language communities in building the technology needed for a voice assistant. This is powerful, as what formerly required experts can now be achieved by small groups of dedicated laypersons using our tools.These technologies come together in the Mycroft devices sold on Kickstarter as the initial reference hardware implementations. Ubuntu Linux packages are now available, and an Android implementation is actively being developed. Implementing a voice control system on virtually any modern single-board computer -- such as the Raspberry Pi -- is now trivially easy.

I realized some key technological pieces were finally reliable and cost effective enough to support the system I wanted to build in the late 90s when working with Microsoft's speech recognition on the first Tablet PC. I began validating my theories and building an "always on" voice system command system. My initial goal was to make voice interaction cheap and easy, allow the democratization of technology. In other words, I wanted to make something easy enough for my grandmother to use -- a system that didn't feel like a computer. Thus began "Christopher". (Yes, he predated the announcement of Alexa by a few months.)Christopher includes: * Free-form, natural language parsing and conversational interaction * Awareness of multiple forms of conversational context * Resolution of overlapping command vocabularies * Whole-house integration and location awareness * User presence awareness * Multiple user support * Screen control/interaction * Easy plug-in developmentInitial plug-ins provided: * Cooking assistance (timers, conversions) * Recipe browsing and shopping list * Reminders and alarms * Google calendar inquiries * IoT / home lighting control * Music jukebox * Weather conditions and forecasts * Wikipedia lookups, including Wikidata support (e.g. "Where was Johnny Depp born?")The system was developed in stealth mode until meeting the Mycroft team...

I led the technical side of the company as we created the most widely successful adaptive traffic (as in cars) control system in the world. We built a $15+M company from scratch and revived an industry that had all but abandoned new technology for the past 30 years.The InSync system uses cutting-edge technology in a way that has been accepted in an industry where there was still debate about whether solid-state relays could safely replace electromechanical relays in traffic cabinets. We introduced them to TCP/IP networking, image analysis, distributed computing and real-time peer-to-peer traffic optimization. Turning naysayers into our advocates was fun, but the best part was improving the environment, reducing crashes, and giving people back a few minutes of their lives every day by improving the efficiency of traffic flow.

Worked with an incredibly talented group of people building the VisualCell platform, used to model and simulate complex biological systems such as cancer cells and perform in silico experiments. This evolved into the REFS (Reverse Engineering, Forward Simulation) platform for discovering probabilistic causual networks from experimental data.Also created tools for the VisualHeart platform, a National Institutes of Health funded project to research the propagation of electrical signals through cardiac tissue, attempting to determine the mechanism and base causes of ventricular fibrillation. This was done in collaboration with IBM as one of the first users of their Blue Gene supercomputer -- the most powerful in the world at the time.

Continued to work on the Drafix product line under the new name of AutoSketch after the Softdesk acquisition. Lead the development team for the final major version.Also worked on early mobile devices (Pocket PC) for viewing and markup of CAD drawings. Then joined the GIS group to work on early digital mapping technologies including an application using the then-new touch, gesture and voice-control capabilities of Microsoft's Tablet PC, which was featured onstage at Microsoft's product launch.

Worked on the core technology behind the Drafix and Planix product lines. I led the effort to move the product from C and the early Windows SDK style programming into C++ and object oriented models.