Christophe @ work

Almost 6 years at Google in the bay area already, in the Global Networking organization, building telemetry for on one of the largest networks in the world.

I have spent the last 30 years analyzing internet data and building tools to improve network performance and availability. I am also an active contributor to the research community (through ACM SIGCOMM) and I have been an advocate of making data publically available for researchers (in particular as M-LAB executive sponsor in Google -- M-LAB is a non profit organization funded in part by Google).

I am deeply concerned by the craziness around ML or AI which is becoming the hammer to fix anything (in networking in particular). I do not consider myself an expert, but an educated user of statistical learning and someone who deeply understand networking data and how to use them for the benefits of all.

Bio

I have a PhD in Computer Science (INP Grenoble, FRANCE, January 1991) and an HDR (INP Grenoble, September 1996). I joined INRIA Sophia Antipolis in October 1993 where I worked with Christian Huitema and Walid Dabbous on internet protocols and applications. This is when I started working on multicast, and designed a peer-to-peer multicast game on the Mbone (fully peer-to-peer). At INRIA, I also worked on automatic protocol generation from ESTEREL formal descriptions. From October 1998 to April 2003, I worked at Sprint Advanced Technology Labs in San Francisco. I was given the opportunity (thanks Franck de Nap and Bryan Lyles) to build a new research group which task was to understand performance of the Internet and help with backbone engineering and design. Outstanding experience with amazing people. We designed a unique measurement infrastructure made of GPS synchronized packet collection, routing information, SNMP data, and active probing. Everything had to be invented at the time. At Sprint, we also did seminal work on DiffServ together with Martin May and Jon Crowcroft. In 2003, I moved to INTEL Research in Cambridge where the objective was to create a multi-disciplinary research group to help INTEL identify and develop promising new technologies. This is where we started Pocket Switched Networks (also known as "Haggle"), and designed new anomaly detection techniques for the Internet, among other cool projects. I joined THOMSON in Paris in October 2005 to launch a  research lab on new communication services and platforms for contents delivery. The Paris lab pioneered research on home networking, home (nano) data centers and CPE added-value features in general. From July 2008 to April 2009, I acted as the CTO of Thomson Corporate Research. In May 2009, I became Chief Scientist for Thomson, in charge of long term innovation strategy and relationship with academics (among other stuff ). In January 2010, THOMSON became TECHNICOLOR.  In 2011, I created a research lab in Palo Alto. The Palo Alto lab  was performing research on user understanding, privacy, and recommendations. I joined SAFRAN in September 2015 as CTO of SAFRAN Analytics, a new entity created by SAFRAN to develop a data culture in the group and to create value from data collected in group companies. This was quite a different challenge for me as SAFRAN Analytics is a business entity, not a research or innovation lab. I was also involved with SAFRAN TECH, the SAFRAN R&T organisation. Since january 2017, I have been appointed SAFRAN "emeritus expert", the highest level of technical expertise inside SAFRAN. 

In June 2018, I moved back to California to join Google, where I have built a telemetry team focused on statistical analysis of network data. We have designed and deployed two  innovative products:

• We co-invented the Network Intelligence Center which is the first customer facing data analytics tool in the Cloud (according to Gartner). We designed accurate loss and latency dashboards  (real-time) using statistical inference (mathematically proven 2% error).

• More recently, we launched a real-time global network anomaly detection tool that automatically mitigates network problem that we detect and localize automatically, relying on simple and scalable statistical analysis. We manage to spend less than 6mns between detection and automated mitigation, which is also unique in industry.

According to Google Scholar, my H index is 84. I am an ACM Fellow since 2005.

Curriculum

My 2024 Resume is available on request at christophe.diot @ gmail.com. Some patents are missing, a few publications. Google Scholar has more accurate information. For the older stuff I should have a paper copy somewhere.

Publications, Reports, Patents

My Publications and reports are available online. One day, I will find enough time to re-organize it by topic, but that's really a lot of work :-(