Charilaos Mylonas

Charilaos Mylonas

Computational Scientist

Deloitte, Risk Advisory


I am working as a senior consultant at Deloitte within the Risk, Data, and Analytics team of Risk Advisory.

I have extensive experience using deep learning and statistical techniques through personal, academic, and work-related projects, covering a wide variety of data, such as high-frequency signals (e.g., sensor data/simulation data/speech data), SCADA data (Wind Farm monitoring), financial data (e.g., financial risk computations), health science related data (biological signals, skin lesion images, knowledge graphs (e.g., DrugBank), protein/small molecule embeddings). Knowledge graphs and efficient large language model training are my (main) current passions!

My main expertise is on (Message-passing) Graph Neural Networks (1), Bayesian Deep Learning for Generative modeling (e.g., (3, 4) ) and their fusion (e.g., such as VAEs for graph structured data 5, and Bayesian predictive models for time-series on graphs 6). I have also developed Bayesian deep learning models for “irregular” time-series occuring in Structural Health Monitoring (SHM) (7), and co-authored papers on using Neural ODEs for SHM (8).

These are some of the techniques I employed during my PhD at ETH Zurich.

Download my resumé.

  • Probabilistic deep learning & Deep Generative Models
  • Large Language Models and efficiency (e.g., qLoRA) / LLM agent applications
  • Graph Neural Networks
  • Financial Risk Modeling
  • Scientific Computing and Optimization
  • PhD Machine Learning for SHM under Uncertainty, 2021

    ETH Zurich

  • MSc Computational Science & Engineering, 2015

    ETH Zurich

  • MSc in Civil Engineering, 2012

    Aristotle University of Thessaloniki

Recent Publications

Quickly discover relevant content by filtering publications.
(2021). Foundations of population-based SHM, Part IV: The geometry of spaces of structures and their feature spaces.

PDF Cite

(2021). Structural identification with physics-informed neural ordinary differential equations. NeuralODEs application.


(2019). UQLab user manual--Sensitivity analysis.. UQLab Sensitivity analysis manual.

PDF Cite

(2019). UQLab User Manual—Canonical Low-Rank Approximations.

PDF Cite


Senior Consultant
Feb 2022 – Present Zurich
Implementation of advanced analytics prototypes, DevOps, and cloud computing
Doctoral Researcher
ETH Zurich
Sep 2016 – Nov 2021 Zurich

Chair of Structural Mechanics and Monitoring. Research topics:

  • Generative Models for UQ in engineering
  • Graph Neural Networks
  • Wind turbine and wind farm simulations
Research Assistant
ETH Zurich
Dec 2015 – Aug 2016 Zurich
Scientific Software Developer, Chair of Risk, Safety, and Uncertainty Quantification
(MSc thesis writing)
ETH Zurich
Dec 2014 – Aug 2015
Shape optimization with Boundary Elements
Investment Banking Internship / Full-Stack Software Engineer
Credit Suisse
Jul 2014 – Dec 2014
Implemented from scratch in Javascript and Python internal web-based tools for time series inspection (e.g., trading signal discovery), implemented a R-to-C++ interface for an option pricer.