About自己紹介
I build machine-learning systems that pick out faint signals in noisy data. Most of the work is the engineering that makes them trustworthy: validating them with clinicians against clinical reference standards, and keeping them reliable once they are in clinical use.
Most of my work is on signals that are difficult to read. They are faint, they sit in a lot of noise, and they feed decisions where a wrong answer costs something.
I did a PhD in computational neuroscience at King’s College London, on deep learning and explainable AI for brain connectivity. Before and around that I spent more than eight years in industry, on the Linux kernel, embedded systems, and machine-learning platforms. I also co-founded a health-tech startup and looked after its technology. The common thread was getting a promising idea to work reliably once it left the notebook.
These days I work as a Research Software Engineer in the ECG Core Lab at the University of Glasgow, with Prof. Peter Macfarlane and Prof. Derek Connelly. The lab has developed the Glasgow ECG analysis program for more than fifty years, and it now helps read over 20 million recordings a year. My main project is a deep-learning layer that improves how the program detects atrial fibrillation. It reads the faint atrial part of the signal using weak-signal methods from astronomy and geophysics, and it strengthens the existing algorithm instead of replacing it. The code is used in clinics, so a mistake has real consequences, and we build it that way.
Day to day I work on detection in noise, time-series modelling, denoising, and anomaly detection. Astronomy and geophysics have spent decades refining the same methods to pull faint sources out of a loud background, and some of that thinking is already in the ECG work. I want to keep going in that direction, toward larger data and fainter signals.
Toolbox道具
The languages, frameworks, and systems I reach for.
Languages
ML / DL
Signal & data
Systems & scale
Domains
Always
learning現在進行形
Experience経歴
Sixteen years across academia and industry, in research and engineering.
Research Software Engineer
University of Glasgow · Glasgow, UK
- Working in the ECG Core Lab led by Prof. Peter Macfarlane and Prof. Derek Connelly, on the Glasgow ECG analysis program. The program is clinically deployed and helps interpret more than 20 million electrocardiograms a year.
- Building a commercial deep-learning layer that improves atrial-fibrillation detection. It reads the faint atrial signal with methods borrowed from astronomy and geophysics.
- Validating against expert-adjudicated reference standards and external databases, developing to IEC 60601-2-51 under an ISO 9001:2015 quality system.
- Adding modern ML practice (rigorous evaluation, testing, reproducibility) to a codebase the lab has developed for over fifty years, so it stays dependable in clinical use.
Co-founder & CTO
Tycho MedLink · London, UK
- Co-founded a digital-therapeutics startup and led the technology for a VR treatment for Seasonal Affective Disorder, from prototype to clinical pilot.
- Raised £100k across two UKRI rounds through the UCL and Cambridge alumni accelerators. Ran an early pilot with UCL Hospitals with promising results.
- Built the VR product in Unity for Meta hardware, with instrumented user metrics.
PhD Researcher, Machine Learning & Neuroimaging
King's College London · London, UK
- Developed graph neural networks with explainable-AI attribution (SmoothGrad, Grad-CAM) to predict neurodevelopmental outcomes from neonatal brain connectivity.
- Worked in the CoDe Neuro lab alongside clinicians at the Centre for the Developing Brain. Results were published and presented at OHBM.
Senior Research Engineer
Prepaire · Dubai, UAE (Remote)
- Led development of AI models, including custom LLMs, that automated genomic data-analysis pipelines and cut processing time by around 40%.
- Shipped the models into a high-performance production environment built to scale.
Teaching Assistant, Deep Learning
Neuromatch · Remote
- Taught deep learning with PyTorch and neuroimaging tooling to an international cohort. Ran daily labs and project work.
Research Software Engineer
Google Summer of Code · London, UK
- Built an infant eye-tracking API prototype, mentored by McGill University’s ophthalmology group.
Technical Editor, Computer Vision
RSIP Vision · Remote
- Reviewed and summarised new computer-vision and medical-imaging research for Computer Vision News.
Senior Machine Learning Engineer
Saddington Baynes · London, UK
- Built AI image-processing automation in TensorFlow that cut manual work by about half. Set up GPU-accelerated Docker and CI/CD for model deployment.
Software Engineer, Linux Kernel
Microsoft · UK
- Worked on cloud hypervisor performance and stability at the kernel level, in C. Contributed to virtualization R&D.
Systems Software Engineer
Kano Computing · London, UK
- Built and maintained a Linux-based OS, including system services and Qt/C++ and GTK applications. Cut the image build time from 4 hours to 30 minutes and added CI/CD.
Research Software Engineer, Serious Games
University of Athens · Athens, Greece
- Built an accessible “serious game” for children with mild disabilities (Epinoisi R&D), in PyGame and WebGL with a C++ game AI.
Embedded Systems Engineer
INTRACOM Defense Electronics · Athens, Greece
- Embedded R&D (FPGA, microcontrollers) for a military communications system under NATO clearance. Built an automated test framework validated to NATO and MIL-STD requirements.
Publications論文
Peer-reviewed papers, conference work, and abstracts.
μ-Opioid Modulation of Sensorimotor Functional Connectivity in Autism: Insights from a Pharmacological Neuroimaging Investigation using Tianeptine
Dimitrov, M., Wong, N.M.L., Leaman, S., França, L.G.S., Valasakis, I., He, J., Lythgoe, D.J., Findon, J.L., Wichers, R.H., Stoencheva, V., Robertson, D.M., Blainey, S., Ivin, G., Holiga, Š., Tricklebank, M.D., Batalle, D., Murphy, D.G.M., McAlonan, G.M., Daly, E.
Biological Psychiatry Global Open Science
Explainable Deep Learning for Subtyping: A SmoothGrad Approach
Valasakis, I., Batalle, D., Deprez, M.
OHBM 2024
Predicting Neurodevelopmental Phenotypes from Neonatal Brain Connectivity using Graph Neural Networks
Valasakis, I., Batalle, D., Deprez, M., McAlonan, G.
OHBM 2023
Deep learning-based reconstruction for 3D coronary MR angiography with a 3D variational neural network (3D-VNN)
Qi, H., Hammernik, K., Lima da Cruz, G., Valasakis, I., Rueckert, D., Prieto, C., Botnar, R.
ISMRM 2021
Development of a Processing Toolset for Ion Mobility Mass Spectrometry
Valasakis, I.
MSc Thesis, Birkbeck, University of London
Contact連絡
Happy to talk about interesting problems in signals and machine learning. Email is the fastest way to reach me.
Open to
- Research and engineering roles in signals and machine learning
- Collaborations on time-series and detection problems
- Speaking, peer review, and technical writing