Georgios Damaskinos

Georgios Damaskinos

I am a machine learning engineer at Meta London focusing on recommender systems. I received my Ph.D. from École Polytechnique Fédérale de Lausanne (EPFL) where I worked mainly on private and robust machine learning but also on recommender systems. I am mainly a practitioner but also study algorithmic tools from a theoretical perspective.

Interests

  • Machine Learning
  • Recommender systems
  • Privacy
  • Mobile and distributed computing
  • Natural language processing

Education

Experience

 
 
 
 
 

Machine Learning Engineer

Meta (formerly Facebook)

Oct 2020 – Present London, UK

Working as a Machine Learning generalist on multiple areas related to creating and productionizing deep models (data, modeling, offline evaluation, online testing, deployment) in the following fields.

  • [06/2021 - Present] Recommender systems for e-commerce: Working on product ranking for Facebook and Instagram Shops (2021 world cloud summit talk, 2022 global AI conference talk). Language: Python, SQL, C++
  • [11/2020 - 06/2021] Natural language processing for Workplace: Developed solutions for video understanding and personalization (paper [8], 2021 conversational AI & NLP summit panel, 2021 data innovation summit talk and interview). Language: Hack, SQL, Python
 
 
 
 
 

Software Engineering Intern

Meta (formerly Facebook)

Jun 2019 – Aug 2019 London, UK
Worked on the DELF paper [5] regarding the privacy challenges of the deletion infrastructure by mainly focusing on performance evaluation. Implemented optimizations as a result of the visibility that this evaluation provided.
Language: Hack, SQL, Python
 
 
 
 
 

Research Intern

IBM

Jun 2018 – Sep 2018 Zurich, Switzerland
Formally derived and implemented a differentially private algorithm for convex optimization of generalized linear models, as part of the core algorithmic component of SnapML. The work continued at EPFL and led to [7].
Language: Python
 
 
 
 
 

Research Intern

Technicolor

Jun 2017 – Aug 2017 Rennes, France
Designed and implemented a method to boost the evaluation of a new recommendation service by exploiting the output of a supposedly well-established online recommender, acting as a black-box that the new service can query. Published patent PCT/EP19/052345. The work continued at EPFL and led to [4].
Language: Python
 
 
 
 
 

Web Developer

Athens Clue

Dec 2014 – Dec 2014 Athens, Greece
Designed and implemented a web application for an escape room.
Language: JavaScript, HTML, CSS
 
 
 
 
 

Tutor

High School Students

Mar 2012 – Jun 2015 Athens, Greece
Provided complementary mathematics, physics, and programming lectures to 8 students who either needed additional help or wanted to go beyond their school curriculum and perfect their skills.

Awards

  • Best Paper Award, Middleware 12/2020
  • Best Teaching Assistant Award, EPFL 12/2019
  • Runner-up for Best Internship Award, Technicolor 11/2017
  • EDIC PhD Fellowship, EPFL 04/2015
  • Thomaideio Award for the 3rd best performance among graduates, NTUA 2015
  • Papakyriakopoulos Award for excellence in Mathematics, NTUA 2011

Publications

  1. "Unsupervised Topic Segmentation of Meetings with BERT Embeddings"
    A. Solbiati, K. Heffernan, G. Damaskinos, S. Poddar, S. Modi, J. Cali
    2. Paper Code Video
  2. "Differentially Private Stochastic Coordinate Descent"
    G. Damaskinos, C. Duenner, R. Guerraoui, N. Papandreou, T. Parnell
    AAAI 2021 (Conference on Artificial Intelligence)
    PPML @ NeurIPS 2020 (Conference on Neural Information Processing Systems)
    3. Paper Code PPML Short Video AAAI Long Video Poster
  3. "FLeet: Online Federated Learning via Staleness Awareness and Performance Prediction"
    G. Damaskinos, R. Guerraoui, A.M. Kermarrec, V. Nitu, R. Patra, F. Taiani
    Middleware 2020 (best paper award)
    Transactions on Intelligent Systems and Technology 2022 (Journal)
    4. Paper Code Video
  4. "DELF: Safeguarding deletion correctness in Online Social Networks"
    K. Cohn-Gordon, G. Damaskinos, D. Neto, J. Cordova, B. Reitz, B. Strahs, D. Obenshain, P. Pearce, I. Papagiannis
    USENIX Security Symposium 2020
    5. Paper Video
  5. "The Imitation Game: Algorithm Selection by Exploiting Black-Box Recommenders"
    G. Damaskinos, R. Guerraoui, E. Merrer, C. Neumann
    NETYS 2020 (International Conference on Networked Systems)
    6. Paper Code Video
  6. "AggregaThor: Byzantine Machine Learning via Robust Gradient Aggregation"
    G. Damaskinos, E. El Mhamdi, R. Guerraoui, A. Guirguis, S. Rouault
    MLSys 2019 (Conference on Machine Learning and Systems)
    7. Paper Code Video Poster
  7. "Asynchronous Byzantine Machine Learning (the case of SGD)"
    G. Damaskinos, E. El Mhamdi, R. Guerraoui, R. Patra, M. Taziki
    ICML 2018 (International Conference on Machine Learning)
    8. Paper Code Poster
  8. "Capturing the Moment: Lightweight Similarity Computations"
    G. Damaskinos, R. Guerraoui, R. Patra
    ICDE 2017 (IEEE International Conference on Data Engineering)
    9. Paper Code Poster

Teaching

I have been a teaching assistant for the following EPFL courses:

I have supervised 1 Ph.D. semester project, 1 MSc thesis, 5 MSc and 1 BSc semester projects and 1 intern.

Service

I have been a reviewer for: