Hello, I am Mahmut Ozge Karakaya, Ph.D.
Computer Engineer and Guitar Player

I'm a Computer Engineer. I hold a Ph.D. in Computer Engineering. I've studied Machine Learning and Recommender Systems. I'm working as a Machine Learning Engineer in my current job. I am mainly interested in Machine Learning and Web Technologies. I have a good understanding of algorithms and tools for Machine Learning and Web Applications. I use several programming languages such as Python, Java, Javascript, and Scala. I implement both backend and frontend applications. I have contributed to several web applications that are heavily used by clients on production. I love agile methodologies, and I am a certified Scrum Master. I like sharing. I have represented nine tech-talks so far. I also like automation. I love coding CI/CD pipelines and monitoring. Finally, I am interested in music. I like recording guitar covers with Garage Band and Fender.

Find me on web;

Resume

Go to next/previous page

Timeline of Projects and Achievements

  • A blog post on Medium - The Timeline of Recurrent Neural Networks and How to Use the Improvements in Tensorflow
    2020

    I have published a blog post about Recurrent Neural Networks on Medium. This post explains the important improvements on RNN since it was invented in 1985. You can also find related implementation in Tensorflow for each improvement. The link is here.

  • Defended my Ph.D Thesis - Diversification in Recommender Systems
    2020

    I have defended my Ph.D. Thesis. The topics covered in the thesis;

    • 1) Diversity Evaluation of Session-Based Recurrent Neural Networks

    • 2) Enhancing Diversity with k-Gated RNN - A new RNN cell for more diverse recommendations

    • 3) Objective function modification of Singular Value Decomposition to Increase Aggregate Diversity in Recommender Systems

    • 4) A graph based approach to Increase Aggregate Diversity in Recommender Systems

    • 5) Usefulness Definition for Serendipity in Recommender Systems.

  • Published an academic paper on Springer - Effective methods for increasing aggregate diversity in recommender systems
    2017

    I have published an academic journal about Recommendation Systems. In this journal we introduce a graph based algorithm, and we change optimization function of SVD to improve aggregate diversity of recommendations. You can find my paper here.

  • Founder at HaberFili
    2013 - 2016

    HaberFili is a Turkish news feed website which is a guide to read news. HaberFili collects news from different sources and recommends them to its users regarding their taste. You can also find similar news for each one.

  • Speaker at Ozgur Yazilim ve Linux Gunleri
    2015

    I was one of the speakers at OYLG15. I talked about Apache Mahout. You can find my presentation here.

  • Published an academic paper on Springer - Clustering-based diversity improvement in top-N recommendation
    2014

    I have published an academic journal about Recommendation Systems. In this journal we introduce a new algorithm to improve diversity of recommendations. You can find my paper here.

Certificates

Education

  • Ph.D. Computer Engineering at Bahcesehir University
    2012 - 2020
  • MSc Computer Engineering at Bahcesehir University
    2009 - 2012
  • BSc Computer Engineering at Ege University
    2003 - 2008

Experience

  • Machine Learning Engineer at DigiB
    2020 - Present
  • Dev Engineer at ING Bank
    2017 - 2020
  • Lead Engineer at GE Aviation
    2015 - 2017
  • Senior Software Developer at Sony
    2013 - 2015
  • Software Development Assistant Manager at BNP Paribas Cardif
    2012 - 2013
  • Software Engineer at Anadolu Sigorta
    2008 - 2012
  • Software Engineer at Usta Yazilim
    2007 - 2008

Blog

Go to next/previous page

The Timeline of Recurrent Neural Networks and How to Use the Improvements in Tensorflow

2020

Recurrent Neural Networks (RNNs) process the data as a sequence. Therefore, it is critical in applications where the order of the input provides information, e.g., music composition, speech recognition, handwriting recognition, etc. In this article, we review the timeline of RNNs’ evolution. I note significant improvements in historical order. You can also find the implementations of the contributions in TensorFlow (the version is 1.11.0).

Music

Go to next/previous page

B.B. King Thrill Is Gone GarageBand cover

2015

B.B. King is one of the greatest guitarists of all time. He is also an influencer for many young guitarists. He contributed to blues very much but also many famous guitarist from other genres are inspired by him. When it comes to solo his playing got leaner in time and that made his style perfect with his signature vibratos, slides, limited note usage etc. Here is a cover of one of his greatest songs, Thrill Is Gone, that I've recorded with my guitar and GarageBand after his passing.

Contact info

  • mokarakayaatgmaildotcom