Current Position & Biography
Martijn de Vos is a postdoctoral researcher in EPFL’s Scalable Computing Systems Laboratory in Switzerland. His current academic page emphasizes efficient machine-learning systems, including distributed and decentralized approaches, while retaining distributed-ledger technology among his research interests. EPFL’s directory lists him as a scientist in the same laboratory. He completed his doctorate in 2021 on decentralization and disintermediation in blockchain-based marketplaces, supervised by Dick Epema and Johan Pouwelse.
Notable Contributions
De Vos coauthored XChange: A Blockchain-based Mechanism for Generic Asset Trading In Resource-constrained Environments with Can Umut Ileri and Johan Pouwelse. The work separates trade management from asset exchange, using a five-phase protocol and TrustChain records to support trading across isolated blockchain platforms while limiting adversarial fraud. Its authors evaluated an implementation on low-resource devices and a compute cluster. Reported performance belongs to those experiments; it should not be read as a universal benchmark for production exchanges or a claim that cross-chain fraud has been eliminated.
Why Katharos.ai Is Following Them
Katharos.ai follows de Vos because decentralization depends on the design of individual components, not merely on attaching a ledger to an existing marketplace. Separating discovery, negotiation and settlement can reveal where control still concentrates and which parties must be trusted. His research trajectory also offers a useful connection to distributed AI: both areas require explicit choices about coordination, resource sharing and the consequences of unreliable participants.
Areas to Follow
Peer-to-peer markets, cross-chain asset exchange and decentralized computing. Coverage will examine whether a design reduces reliance on an operator or simply moves that reliance elsewhere. Experimental results should be interpreted alongside the hardware, workload and adversary assumptions used to obtain them. For current machine-learning work, the relevant question is what can be coordinated securely and efficiently without assuming that every participant behaves identically.
Reporting Scope
Switzerland identifies his current EPFL professional base, supported by his academic page and the institution’s Lausanne directory entry; it is not a nationality claim. The blockchain marketplace work discussed here predates his present emphasis on machine-learning systems. The directory’s scientist title and his own postdoctoral description refer to the same laboratory. This profile does not present him as the operator of a cryptocurrency exchange or recommend a token.
Sources for Further Reading
https://devos50.github.io/ https://people.epfl.ch/martijn.devos https://arxiv.org/abs/2004.05046
Photo Credit
Portrait via Martijn de Vos’s personal academic website. Site copyright: © 2026 Martijn de Vos. Photographer not identified in original metadata. Whole photograph downsampled and padded without cropping or facial alteration; original EXIF, XMP and ICC metadata retained. No unrestricted reuse license is asserted.
Editorial Disclaimer
This profile summarizes information from the cited sources. Katharos.ai does not endorse the subject or guarantee every source claim’s authenticity.