Igor Kulev

I am passionate about applying machine learning to real-world problems.

Experience

Data Scientist

Roche

Develop machine learning methods and algorithms to generate impactful evidence and insights for Roche business. Optimize business performance, solve complex data problems and deliver the insight that helps to define Roche's strategy and enable the organization to reach patients faster.

September 2020 - Present

Researcher (Data Analytics)

REACH 2020

Extract and analyze time-series data obtained from fitness trackers. Analyze trial data. Develop novel machine learning methods to recommend personalized interventions. Prepare deliverables, reports and presentations. Organize Data Analytics workshop. Develop a mobile app to promote physical activeness.

February 2016 - January 2020

Doctoral Assistant

Artificial Intelligence Laboratory and Human Computer Interaction Group
École Polytechnique Fédérale de Lausanne

Conduct research on personalizable intervention systems to promote healthy behavior change. Design, implement and evaluate different machine learning models. Participate in teaching courses: Information, Computation and Communication; Intelligent Agents; and Human Computer Interaction. Supervise student projects. Maintain and manage servers.

September 2014 - January 2020

Teaching and Research Assistant

Faculty of Computer Science and Engineering
Ss. Cyril and Methodius University in Skopje

Participate in teaching courses: Algorithms and data structures; Advanced Algorithms; Intelligent User Interfaces; Intelligent Information systems; Multimedia Systems; Computer architecture and organization. Prepare teaching materials. Conduct research on collaborative health-care systems.

September 2011 - August 2014

IT Consultant

Macedonian Civic Education Center

Design and implement a web-based questionnaire with complex questionnaire flow.

August 2012 - September 2012

Summer Student Intern

The European Organization for Nuclear Research (CERN)

Simulate the behavior of antihydrogen atoms under static magnetic and temporary varying electric fields.

June 2010 - August 2010

Education

École Polytechnique Fédérale de Lausanne

PhD degree in Computer Science
School of Computer and Communication Sciences
Machine Learning Recommender Systems Personalization Time Series Clinical Trials Supervised Learning Unsupervised Learning Adversarial Learning Causal Inference
September 2014 - January 2020

Ss. Cyril and Methodius University in Skopje

Master degree in Electrical Engineering and Information Technologies
Program: Intelligent Information Systems
GPA: 10.0
Recommendation Algorithms Collaborative Filtering Personalization Time Series Supervised Learning Health-Care Systems
November 2011 - April 2013

Ss. Cyril and Methodius University in Skopje

Bachelor degree in Electrical Engineering and Information Technologies
Program: Informatics and Computer Engineering
GPA: 10.0
Algorithms Machine Learning Probability and Statistics Software Development
September 2007 - June 2011

Online Courses

AI for Medicine
This specialization teaches how to apply machine learning to concrete problems in medicine.
AI for Medical Treatment
This course teaches how to estimate treatment effects and how to apply machine learning interpretation methods to explain the decision-making of complex machine learning models.
AI for Medical Prognosis
This course teaches how to develop AI models to predict the future health of patients.
AI for Medical Diagnosis
This course teaches how to apply AI to medical imaging to diagnose diseases.
Fundamentals of Machine Learning in Finance
This course aims at helping students to be able to solve practical ML-amenable problems that they may encounter in real-life.
Guided Tour of Machine Learning in Finance
This course aims at providing an introductory and broad overview of the field of ML with the focus on applications on Finance.
The HackerRank Interview Preparation Kit
This kit consists of a set of algorithmic problems organised around core concepts commonly tested during interviews.
Feature Engineering
This micro-course teaches how to extract features from raw data.
Intermediate Machine Learning
This micro-course teaches how to handle missing values, non-numeric values, data leakage and more.
Intro to Machine Learning
This micro-course teaches the core ideas in machine learning.
Pandas
This micro-course teaches how to manipulate and analyze data.
Python
This micro-course teaches Python.
Spark
This course teaches how to use Spark to work with big data and build machine learning models at scale, including how to wrangle and model massive datasets with PySpark.
Game Theory II: Advanced Applications
This course teaches how to design interactions between agents in order to achieve good social outcomes.
Game Theory
This course teaches the basics of game theory: representing games and strategies, the extensive form, Bayesian games, repeated and stochastic games, etc.
Deep Learning
This tutorial teaches the main ideas of unsupervised feature learning and deep learning.
Algorithms: Design and Analysis, Part 1
This course teaches fundamental principles of algorithm design.

Technical Skills

Programming Languages
  • Python
  • Java
  • Matlab
  • SQL
  • C
  • C++
  • C#
  • JavaScript

Tools
  • Tensorflow
  • Scikit-learn
  • Pandas
  • PySpark
  • Flask
  • Amazon Web Services
  • Microsoft Visual Studio
  • Eclipse

Projects

Movie Suggestions New
Developed a web site that recommends movies based on keywords. The app uses machine learning to understand which keywords are related to each other and which movies are related to which keywords.
Recurrent Neural Network Recommender System Information Retrieval Website Crawling Data linking Tensorflow Back-end Flask Java Front-end JavaScript jQuery Deployment Amazon Web Services
Smart Compose
Collected data and developed a sequence-to-sequence model to offer relevant and real-time suggestions as you type. The model was inspired by Gmail Smart Compose. Implemented a fast beam-search algorithm and developed a web site to achieve a real-time inference.
Sequence-to-Sequence Recurrent Neural Network Natural Language Processing Language Modeling Website Crawling Tensorflow Back-end Flask Java Front-end JavaScript jQuery Deployment
HealthyTogether
Developed a machine learning model to predict behavior change under different interventions. Integrated the model into a mobile app to promote physical activeness. The app was used in an experimental study. Collected data and analyzed the results of the study.
Recommender System Back-end Flask Deployment Tensorflow
Reducing Intervention Bias using Adversarial Balancing
Developed a novel adversarial approach to reduce bias when estimating the intervention effect from observational data. Demonstrated that this approach performs better than the existing approaches on a widely-used benchmark dataset.
Causal Inference Observational Data Intervention Bias Adversarial Learning Representation Learning Deep Neural Network Individualized Treatment Effect Tensorflow
Personalizable Intervention System for Senior Adults
Proposed a novel intervention system to promote physical activeness in senior adults. The system uses minute-by-minute step count data to recommend a mobile app intervention that is most likely to work for the target user based on his or her activity patterns.
Recommender System Randomized Trial Supervised Learning Autoencoder Representation Learning Time Series Causal Inference Tensorflow
Discovering Intervention Profiles From Time Series Data
Proposed a novel method to discover and predict behavior change patterns from frequently-sampled sensor data. Demonstrated that the system produces explainable patterns that may be used to recommend strategies for healthy behavior change.
Recommender System Time Series Behavior Patterns Behavior Change Patterns Mixture Model Representation Learning Latent Variable Model Tensorflow Matlab
Stock Market Trend Prediction
Developed a method based on a Recurrent Neural Network and a Mixture Model to group companies based on their stock market movements before and after the stock market crash in 2008. Used the model to predict how individual stocks change their behavior after the crash based on pre-crash time series data.
Mixture Model Latent Variable Model Deep Learning Recurrent Neural Network Clustering Website Crawling Time Series Modeling Regression Behavior Patterns Behavior Change Patterns Tensorflow
Intervention-Based Clustering
Proposed a Bayesian mixture model to identify subpopulations with different behavior changes from longitudinal data. Showed that the model can discover the subpopulations that respond to the intervention from a limited amount of data.
Bayesian Mixture Model Clustering Latent Variable Model Longitudinal Data Randomized Controlled Trial Heterogeneous Treatment Effect Subgroup Analysis Matlab

Selected Publications

Kulev, I., Walk, C., Lu, Y., and Pu, P. Recommender system for responsive engagement of senior adults in daily activities. Journal of Population Ageing (2020).

Kulev, I., Pu, P., and Faltings, B. A Bayesian approach to intervention-based clustering. ACMTrans. Intell. Syst. Technol. 9, 4 (Jan. 2018), 44:1–44:23.

Kulev, I., Pu, P., and Faltings, B. Discovering persuasion profiles using time series data. In Proceedings of the 2nd NIPS Time Series Workshop (2016).

Vlahu-Gjorgievska, E., Koceski, S., Kulev, I., and Trajkovik, V. Connected-health algorithm: Development and evaluation. Journal of Medical Systems 40, 4 (Feb 2016), 109.

Trajkovik, V., Vlahu-Gjorgievska, E., Koceski, S., and Kulev, I. General assisted living system architecture model. In Mobile Networks and Management (Cham, 2015), Springer International Publishing, pp. 329–343.

Interests

Apart from being a data scientist, I like hiking — I enjoy exploring the Swiss mountains and lakes. Also, I like watching mystery, thriller and comedy genre movies.

Achievements And Awards

  • Award "Engineering ring" given by the Engineering Institution of Macedonia. The best engineering student of generation 2010/2011.
  • Award "Gold coin" given by the Ss. Cyril and Methodius University in Skopje. The best student of generation 2010/2011.
  • Graduated with Honours, BSc. Eng., Degree, Faculty of Electrical Engineering and Information Technology, Ss. Cyril and Methodius University in Skopje
  • 4th place and winners of Region 8 (Europe, Middle East and Africa) at IEEEXtreme 6.0 programming competition (2012)
  • 1st place at CodeFu 2012 programming competition
  • 7th place at IEEEXtreme 5.0 programming competition (2011)
  • 1st place at National ACM-ICPC contest for algorithmic programming (2009 and 2010)
  • 13st place at ACM South Eastern Europe Contest 2009, Romania
  • Johnson Controls fellowship (2007-2011)
  • Participation at International Olympiad in Informatics 2007, Croatia
  • 1st place at Macedonian Olympiad in Informatics 2007
  • 3rd prize at National Competition in Mathematics 2006
  • 2nd prize at National Competition in Physics 2005