Mohammed Alloulah

Research, Bell Labs, Cambridge, CB3 0FA, United Kingdom
Find me on LinkedIn, Strava
alloulah [@] outlook [.] com

With 15+ years experience in tech-based R&D, I have envisioned, led, and delivered programmes in multiple industrial & academic settings in areas spanning signal processing, machine learning, and 6G network perception.
Outside work, I am a keen runner and a cyclist.


I'll be in Seoul at IEEE ICC '22 to publicise our ongoing radio-visual learning work.

Two successive program committeess at ENSsys '21/'22 in conjunction with SenSys '21 and ASPLOS '22. Please consider submitting your work.

Python port of mm-wave radar synthesis code useful for GAN training available on my public GitHub.

Looking forward to serve on ENSsys '20 program committee in conjunction with SenSys '20.

Great fun to have participated in panel discussions at ENSsys '19.

Atomic norm-based 1D spectral estimation code available on my public GitLab.

Iterative matrix inversion code available on my public GitLab.


Selected examples by descending chronology:

(1) Self-Supervised Radio-Visual Representation Learning

Learn a radio sensing model without labels by tapping into cross-modal mutual information between radio and vision. State-of-the-art self-supervised learning for 6G sensing.

(2) Deep Learning for Autonomous Inertial Navigation

Enhanced inertial navigation by means of deep learning. In addition to devising neural architecture for time-series inertial signals, the main problem tackled is generalisation through spatial diversity, optimal transport, and domain adaptation.

(3) Energy-Aware Inferences

Idea is to perform best-effort inferencing using a variable instantaneous energy budget. Devised degradable inference using knobs at featurisation-level as well as at neural network-level.

(4) 60GHz Sensing: A Novel Kinetic Channel

Future networks in mm-wave bands can exchange information using a form of vibrational (i.e. kinetic) modulation. Fabricated a new metasurface for coherent modulation + radar sensing techniques.

(5) Contactless Physiological Sensing for Sleep Monitoring

Future WiFi networks can monitor physiological sleep markers in order to provide unobtrusive long-term wellness analytics. Technically, we came up with a new signal subspace tracking formulation for isolating breathing effects from body movements. This idea is a derivative of an earlier, theoretically-minded work that looked into the expansion and shrinkage of the signal subspace as a result of human-scale channel disturbances.

(6) Wireless Electrocardiogram (ECG) Physiological Sensing

Digital health will scale archaic medical practices in order to support ageing populations whilst reducing pressure on resources (monetary and/or physical). I devised a joint compression and wireless coding scheme for robustness and ultra-low power medical-grade operation.

(7) Ultra-Low Power WiFi Modulator for IoT

Traditional direct conversion RF transmitter architecture are not suitable for IoT devices with stringent power budget. A polar transmitter in standard WiFi channels is challenging. We made it work with couple key innovations. Achieved 3x power reduction.
Silicon + several patents

(8) Embedded OFDM PHY

Led work on multiple architecture-aware PHY implementations, part of a terrestrial DVB-T2 chipset. Keywords included inter-symbol interference (ISI) and inter-carrier interference (ICI).
Silicon + several patents

Some public slides

IMULet: A Cloudlet for Inertial Tracking
Scalable Wireless ECG Streaming

Publications and Patents

pdf bibliography