Education
Post-doc, FEUP, Portugal. Information theoretic network coding & Optimal control theory with an application to non-linear partial differential equations, an application to traffic flow, 2009-2011.
Ph.D in EE, specialising in telecommunications, thesis Quality aspects of audio communication, Royal Institute of Technology, (KTH), Sweden, 2009.
M.Sc. in computer science, specialism in system design, Manchester University, UK, thesis non-harmonic partials in synthesis design, 1998.
B.Sc. in Physics and Computer Science, Manchester, specialism in thermodynamics, thesis on colour science, 1997.
Employment
RI.SE, Sweden (formerly SICS). 21 years. Telecommunications, IP networking, new mechanisms for the Internet originally, IP-Multicast, Queuing disciplines, real-time voice. Caching, edge and cloud infrastructure. Mostly done with Ericsson and numerous EU partners.
Big data mining, statistics and machine learning. Road traffic and the process industry. More inline with industries outside of telecommunications.
IBM, Germany, 2 years. At the time (with SICS above) working on a stream protocol, called ST-II. Some help with systems, including AFS.
CISRO, Australia, 2 years. Working with R precursor, called S-PLUS. User satisfaction too.
NVIDIA, UK (formerly Du Pont Pixel systems), 1 year. Microcoding a bitslice architecture for both graphic and image processing algorithms. Single pixel architecture, later a SIMD architecture to process more than 1 pixel per time.
Five major technical accomplishments
Modeling of mobile users with communication capabilities (2015-present). Macroscopic fluid flow models were developed to predict road flow and connectivity on road networks. The population density determines the radio connectivity in sparse conditions, and may even contribute to radio interference in dense situations. We have modeled the density using an approach from physics and engineering known as traffic flow theory. It is an example of ITS with road traffic flow and communications. We have developed analytic+numerical solutions for estimating road densities. Recently, we obtained 12 years of real traffic flow data to validate our
research, more at bada.sics.se.
A networked measurement platform (2016). CheesePI is an Internet measurement infrastructure using Raspberry PIs for quantification of Quality of Experience (QoE) for the home user. It measures network characteristics between PIs placed at selected locations around a country. The goal of the project is to measure the quality in a free, neutral and non-biased manner. Individual PIs store the measurement data in a schemaless database, whilst the central server cheesepi.sics.se stores the aggregated data from each PI for analysis, visualisation and dissemination. CheesePi is in use with one larger operator in Sweden and is being evaluated by the regulator.
A real-time voice handover module (2003). The goal was to create a fully-automated module
for voice handovers based on network measurements. A WiFi-GSM PBX product, we implemented a cross-platform module using measured network parameters and precomputed PESQ scores to infer voice quality. The software is still used in products today.
Implemented BSD kernel-based queuing algorithms, WFQ, W2FQ, Virtual clock, needed for reservation protocols like RSVP. They went into Ericsson’s router product. Algorithms such as these are used in class-based scheduling and traffic control in today’s multiservice IP networks.
In 1996 contributed to the design and implementation of a global file system for HP, called JetFile. It was based on Scalable Reliable Multicast. The system was (and still is) well ahead of its time, especially with file systems such as HDFS being available in a cluster only, whereas JetFile worked worldwide (akin to AFS). We used knowledge of multicast (many to many) mechanisms plus scaling issues, building a reliable file system on an unreliable (UDP) network protocol.