Thomas Asikis

Stampfenbachstrasse 48, 8092, Zürich, ·

Research Assistant and PhD candidate designing, implementing and evaluating AI systems that support society. Strong research interest and years of experience in of machine learning and social sciences.

Current Research

ASSET Consumerism Project

ETH is a research partner in the EU funded project "ASSET Consumerism", which is a part of the Horizons 2020 program. My research includes the design, implementation and testing of the recommender system and any machine learning application that supports the analysis and preprocessing of consumer and product data. In the scope of the project, I participated in the organization of the ASSET challenge in the Social Impact Data Hack. ETH proposed the ASSET Challenge about extracting meta-information from unstructured crowd-sourced product data that can be used to make more sustainable purchase decision.

May 2016 - Present

Multi-agent Reinforcement Learning for Collective Intelligence

A collaborative research with researchers from NTUA. The aim of the project is to design and apply a multi-agent reinforcement learning system on simulations and experiments derived from social sciences.

April 2018 - Present

Academic Activity

Publications & Theses

  • Decentralized Collective Learning for Self-managed Sharing Economies, Evangelos Pournaras, Peter Pilgerstorfer & Thomas Asikis
    This paper envisions an alternative unsupervised and decentralized collective learning approach that preserves privacy, autonomy and participation of multi-agent systems self-organized into a hierarchical tree structure. Remote interactions orchestrate a highly efficient process for decentralized collective learning.
    ACM Transactions on Autonomous and Adaptive Systems, 2018
  • Optimization of Privacy-Utility Trade-offs under Informational Self-determination, , Thomas Asikis & Evangelos Pournaras,
    A framework that privatizes user data with differential privacy in a dynamic manner by adding noise. The addition of noise introduces loss of information for stakeholders that want to process aggregate user data. The proposed framework minimizes the information loss by searching for the optimal noise generating parameters that preserve privacy.
    ACM Transactions on Autonomous and Adaptive Systems, 2018
  • Master Thesis on " Part of Speech Tagging in Greek using Word Embeddings and Deep Neural Networks", Thomas Asikis, Supervision Ion Androoutsoupolos
    My master thesis project aimed to improve part of speech tagging accuracy for the Greek language. During the project, several ML models were evaluated, such as Conditional Random Fields, Hidden Markov Models, Long-Short Term Memory Neural Networks and Quasi-Newton Entropy Classifiers. The thesis contributions were state-of-art performing Models and also the creation of a new tagged dataset. The code is open-sourced on Github.
  • Operations Research and Recommender Systems, Thomas Asikis & George Lekakos
    A recommender system based on a meta-heuristic optimization algorithm, which optimizes existing recommendation to achieve higher precision and recall. At the time, the system achieved competitive performance with the widely used collaborative filtering algorithms.
    Proceedings of Human-Computer Interaction, 2014

Presentations, Conferences and Teaching

  • Presentation:EPOS Code Tutorial ETH Course: Self-Organizing Multi-Agent Systems
  • Presentation: Multi-Agent Reinforcement Learning ETH Course: Self-Organizing Multi-Agent Systems
  • Presentation: Train Global, Test Local: Privacy-preserving Learning of Cost-effectiveness in Decentralized Systems
    Jovan Nikolić, Marcel Schöngens, Evangelos Pournaras,
    10th International Conference on Intelligent Networking and Collaborative Systems INCoS, 2018,
  • Code Artifact: Self-regulatory Consumption via Self-determined Personalized Ratings , Thomas Asikis,
    11th IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO), 2017,
  • Seminar programming tutorial in the AUEB Deep Learning Seminar:
    "Practical Deep Learning using Keras, Theano and Deeplearning4j", 2016
  • Teaching Assistant in AUEB B.Sc. course: "Analytics and Personalization Technologies", 2014


Scientific Assistant

Specializing in the usage of learning and optimization algorithms for human behavior modeling and prediction. My daily duties include the design, development and deployment of machine learning models on the ETH cluster. Furthermore, I am responsible for the analysis of experimental results and the (co)authorship of scientific papers, project reports, teaching and proposals.

May 2016 - Present

Data Scientist

Incelligent specializes in cellular network optimization and analysis. I focused mainly on applying big data analysis and machine learning on data of several cellular network operators. I contributed in the development of several machine learning algorithms and pipelines used in the company and introduced the usage of deep neural networks such as LSTMs and Convolutional neural networks. Some of my daily tasks were:

  • Design and implementation of optimization algorithms.
  • Design and implementation of prediction models, especially with deep learning.
  • Usage of big data frameworks for preprocessing and storage of network data.
  • Collaboration with software developers for the development of intuitive UIs and data visualizations.

May 2015 - April 2016

Full Stack Developer

Sleed is a Digital Marketing and E-business company. As a full-stack developer, I took part in many projects involving the development of web services, web sites and e-shops. During my time at Sleed I developed strong web development and database skills.

February 2014 - October 2014

CRM Intern

Daily tasks included assisting in the following:

  • CRM platform management.
  • Management Facebook and Twitter company profiles.
  • Composition of quantitative reports.
  • Analysis of customer feedback.
  • Analysis and reporting for SEO and SEM strategies.

May 2015 - April 2016


PhD Student in Computational Social Sciences

Indicative passed coursers: Deep Learning, Advanced Topics in Information Retrieval and Natural Language Processing, Data Science in Techno-Socio-Economic Systems
Thesis Title (TBD): "Supporting Sustainable Development via Artificial Intelligence"

May 2016 - Present

M.Sc. in Information Systems

GPA: 8.69/10
Indicative coursers: Information Retrieval, Machine Learning and Data Mining, Natural Language Processing
Thesis: "Part of speech tagging in Greek texts with word embeddings and deep neural networks"

September 2014 - April 2016

Erasmus Studies

I joined the Erasmus program during the last year of my B.Sc. I attended several courses on business analytics and computational intelligence. It was during these courses that I came across Machine Learning and studied neural networks for the first time, which later became one of my main research interests.
GPA: 4.17/5
Indicative Courses: Data Mining and Text Mining, Computational Intelligence

August 2012 - January 2013

B.Sc. in Management Science and Technology

The bachelors in Management Science offered me a broad view on several fields such as Computer Science, Economics, Finance, Quantitative Analysis, Management etc. During that time, my initial research interest in optimization and machine learning was developed. Furthermore, I acquired valuable knowledge and several skills that supported me later in pursuing my research interests.
GPA: 8.35/10
Indicative Courses: Mathematics, Information and Telecommunication Systems, Decision Making, Information Systems and Dabases, Advanced Topics on Software Engineering, Statistics, Quantitive Methods in Finance, Digital Content Management and Human Computer Interaction, Decision Making
Thesis (Elective): "Operations Research and Recommender Systems"

September 2009 - September 2013


Domain Knowledge
Optimization, Machine Learning, Behavior Analysis, Recommender Systems, Sustainability, Big Data, Privacy, Distributed Systems
Technical Skills
Programming languages
Java, Python, SQL, R, PHP
Machine Learning & Statistics
deeplearning4j, keras, tensorflow, theano, numpy, scipy, colt, librec, mallet, Stanford NLP, gensim, nltk
Data processing and storage
pandas, MySQL, Hadoop, Apache Spark, SAP Hana
Presentation & Reporting
latex, html5, css, javascript, plotly, holoviews, d3, echarts, Microsoft Office, Jupyter, Worpress


  • M.Sc. Scholarship: Awarded with a scholarship for scoring 4th in the total ranking and 1st in the "Big Data" specialization of the Information Systems master's program.
  • Erasmus. Scholarship: Awarded with a scholarship for passing all courses with a high GPA.
  • GPA within top 15% in B. Sc.