Sep 2016 – Aug 2020 · 3 yrs 11 mos

• I provide technical leadership for the Simulation and Infrastructure teams. Simulation is an important part of the development cycle for self driving systems.
• For Simulation I led the effort to architect for much larger scale, better insights, cost controls, GPU management, and migrating to a hybrid cloud. On the Infrastructure side, I designed and managed ATG's migration to Docker and cloud-based services to replace shared filesystems.
• More recently I've been working on a project to measure correct driving behavior from the self driving software.

Sep 2014 – Sep 2016 · 2 yrs

• I help teams design and implement scalable and highly available backend systems.
• My first project was to help with the design of the new Dispatch platform, which is primarily written in node.js. Through this effort, we developed several new components for building and maintaining distributed systems. Most of these components are open sourced.
• Next I worked on a RPC-based service discovery and routing system that we would now call a "service mesh". This system had a number of novel features at the time like a very efficient forwarding path and lovingly optimized implementations in multiple languages.
• As Uber scaled the Engineering team, it became a lot harder to for developers to work with the fast evolving code base and constant stream of new colleagues. To help make this easier, I worked on a storage system that let developers safely work with production data from their workstations.
• I talked about these systems in some public talks, one of which is linked below.

Jun 2007 – Sep 2014 · 7 yrs 3 mos · San Francisco

• Our goal when building Voxer was to adapt voice communications to the social etiquette of mobile phones. Now that we have abundant connectivity, the traditional model of phone calls and voice mail no longer makes sense.
• At Voxer I designed and helped build the software and services behind a significant player in the mobile messaging ecosystem. Behind the scenes, Voxer is an all data, time-shifting voice communication system that is both live and asynchronous. Data is saved on the sending device, the server, and the receiving device. This design allows users to be as live as they like or as the network will allow.
• The backend infrastructure is a horizontally scalable, service oriented architecture written entirely in node.js. For years now it has been the largest and most complex production deployment of node.js. Scaling Voxer involved pushing and overcoming the limits of node.js, Redis, and Riak.
• We have been granted over 60 patents, both in the US and internationally, all related to the original innovation of two-way, arbitrarily time shifted voice communication.

Feb 2004 – Apr 2006 · 2 yrs 2 mos

• Triple Canopy is a security company with locations all over the world. Many of these locations had no connectivity or other IT infrastructure. My role at Triple Canopy was to design and deploy a global IT infrastructure of secure voice and data communications to support both domestic and international operations.
• Interesting things include:
• build and tune Voice over IP systems using low cost satellite links
• troubleshoot networks and computers in remote locations that I've never visited, by talking to people that I've never met
• manage network security for systems that are under both digital and physical threat
• develop realtime network metrics visualizations for unique conditions of this network

Jan 1998 – Jun 2007 · 9 yrs 5 mos · San Francisco Bay Area

• Between various other full time jobs, I did consulting work for companies building networked applications and systems. Most of this work was covered by non-disclosure agreements, and many of the companies have since changed ownership.
• There were common themes among all of these engagements:
• optimizing response time for browser-based applications
• measuring and understanding user behavior
• developing high availability solutions for a variety of environments

Jan 1996 – Jan 1998 · 2 yrs · San Francisco Bay Area

• The company I worked for was GeoNet, which was acquired by Level(3) while I was an employee. GeoNet (AS3356) had full peering with no transit providers.
• While at GeoNet, I:
• Managed Cisco routers and routing protocols at 4 public exchange points, including BGP filtering, monitoring, and negotiating peering arrangements,
• Coordinated administration of all production Unix servers providing DNS, SMTP, NNTP, and web proxy services,
• Developed web and database tools for Network Operations Center automation and streamlining of customer support,
• Developed one of the first monitoring and accounting systems based on Cisco's NetFlow export. This was used for customer billing, security, and peering relationships,
• Provided second level escalation for engineering and NOC staff.

Jan 1995 – Jan 1996 · 1 yr · Menlo Park, CA

• The company I worked for was EIT, which was later bought by VeriFone, which was later bought by HP, which was later sold off as something still called VeriFone.
• While at EIT, I:
• Managed the 6-member Systems Administration group for the Internet Commerce Division,
• Designed and implemented all levels of the network infrastructure,
• Managed dual-homed Internet connectivity including router management, ISP/telco negotiation, network security, and firewall implementation,
• Migrated computing environment from Macintosh-based systems to Windows NT desktops and Unix servers.