​​Evangelos Eleftheriou and team​

While consumer technology has shifted toward the cloud, archiving data requires cost-effective solutions. Magnetic tape serves this purpose, but as data is packed more densely, signals begin to interfere with one another, creating noise. Early in his career, Eleftheriou originated the concept of Noise Predictive Maximum Likelihood (NPML) detection, extending earlier detection frameworks to explicitly model and predict noise, allowing systems to extract genuine data from noise and interference with significantly improved accuracy at high storage densities.

When IBM exited hard-disk manufacturing in 2002, Eleftheriou led a team that transferred its expertise to magnetic tape. "Magnetic tape has become a critical backbone of large-scale data archiving, widely used by cloud providers due to its cost efficiency and energy advantages."

To enhance reliability, the team developed several techniques to maintain positional accuracy even on degraded media. They integrated error-correcting mechanisms directly into the positional data (LPOS), improving robustness without significantly increasing overhead. Additionally, they introduced additional redundancy and coding techniques at the signal level, enabling systems to recover rapidly from "burst errors" caused by physical defects. To improve data retrieval in noisy environments, they introduced soft-decision word synchronisation, which distinguishes genuine markers from background noise far more effectively than traditional binary methods.

Decades of contributions to data storage technologies

Born in Aliveri, Evia, Eleftheriou grew up in Greece and pursued electrical engineering there. When his family could not afford to fund postgraduate study abroad, a professor helped him secure a scholarship to continue his studies in Canada. He completed both his master's degree and PhD at Carleton University in Ottawa, graduating in 1985. At graduation, his supervisor offered advice that would prove formative: go into industry first, broaden your horizons and only return to academia later if you choose. Eleftheriou followed that guidance and never looked back. Though offered a position in the US, he chose IBM Research – Zurich and has remained in Switzerland ever since.

During his tenure at IBM, Eleftheriou’s work spanned the full spectrum of digital storage, including magnetic recording, flash memory and phase-change memory. He led major strategic shifts, such as contributing to system-level innovations that improved the reliability and enterprise adoption of flash storage and developing system architectures that separate background maintenance operations from user-facing workloads to ensure consistent performance.

His later research shifted focus from basic data storage and memory technologies to in-memory computing, demonstrating that memory structures can be used to perform certain computations in addition to storing data. This work contributed to the broader direction that eventually led to his transition from researcher to entrepreneur. In 2021, he co-founded Axelera AI, a company developing high-performance chips designed to make AI applications more efficient and sustainable.

While at IBM, Eleftheriou’s work contributed to a large patent portfolio used primarily for licensing and cross-licensing. As CTO of Axelera AI, he leads the company’s technology strategy while also overseeing its patent portfolio, working with a small internal team and external European patent law firms to evaluate invention disclosures and protect core technologies.

About the inventor

Evangelos Eleftheriou spent over 35 years at IBM Research, where he rose to the rank of IBM Fellow, the company’s highest technical honour. He is also an IEEE Fellow and a Foreign Member of the US National Academy of Engineering. Over his career, he has authored more than 350 publications and holds over 150 patents.