I am a senior researcher with a focus on Distributed Systems, Artificial Intelligence and Space Computing.
Short Bio
I received my PhD in Computer and Systems Sciences from KTH in 2008 on "Statistical Methods for Computational Markets: Proportional Share Market Prediction and Admission Control". Prior to that I worked as a distributed systems consultant and developer at IONA Technologies and Argonne National Labs. After my PhD I joined HP Labs and continued work on distributed systems while also shifting focus to work on machine learning applications. I was a visiting researcher at Lund University and KAIST working on IoT and AI appliances respectively. After HP Labs I joined Dell/EMC to work on a cloud-native file system product, and later joined a startup on IoT Wi-Fi sensing. Prior to joining RISE I was a researcher at CableLabs, where I investigated AI applications in wireless communication networks.
[ LinkedIn - DBLP - ACM - Google Scholar ]
Selected Papers
Distributed Systems
- SAFE: Secure Aggregation with Failover and Encryption
- Smart Contracts for Mobile Network Operator Bandwidth Sharing
- Notes on Cloud computing principles
- MapReduce optimization using regulated dynamic prioritization
- A statistical approach to risk mitigation in computational markets
AI/ML
- Semantic Navigation for AI-assisted Ideation
- NetGen: A Network Digital Twin Generator
- Evaluation of HetNet Offloading with a GAN-Driven Wireless Testbed
- Global budgets for local recommendations
Space Computing
- SkyMemory: A LEO Edge Cache for Transformer Inference Optimization and Scale Out
- A Cloud in the Sky: Geo-Aware On-board Data Services for LEO Satellites
Active Projects
LeoDOS: A Distributed Operating System for LEO mega-constellations
Mobile broadband communication and earth observation use cases have pushed the boundaries of space-based computing over recent years. Low-Earth-Orbit mega-constellations, thousands of inter-linked satellites capable of providing low-latency, ubiquitous connectivity and high-resolution observations across the surface of Earth, have emerged as viable businesses for space operators such as Starlink/SpaceX, Amazon Kuiper and OneWeb. The operating cost is driven down by improved rocket reuse (SpaceX), network protocol standardization (3GPP NTN), and more advanced antenna technology (phased-array antennas). Combined with the ever-increasing trend of compute capacity scaling and computer miniaturization these developments have resulted in an exponential growth in number of LEO satellite launches over recent years, from 79 to 2113 yearly launches in the decade between 2012-2022. As of May 30th 2025, there were 7578 Starlink satellites alone circulating in LEO constellations. Although dedicated spectrum (S, C, Ku, Ka) is available for these satellites to communicate with ground stations and existing cellular bands are reused for uplinks (e.g. LTE 25 T-Mobile/Starlink), the downlink capacity, i.e. spectrum range available, is a limiting factor in many use cases, e.g. wild fire prevention and detection, that require large volumes of imagery to be collected and sent to ground stations for processing. The problem we address in this project is how we can move some of this processing on-board satellites to reduce the volume of data sent to ground stations.
Funded by the Swedish National Space Agency under contract Dnr 2025-00306.