Andrija Djurisic, Machine Learning Developer in Belgrade, Serbia
Andrija Djurisic

Machine Learning Developer in Belgrade, Serbia

Member since December 11, 2016
Andrija is a talented software engineer and a machine learning researcher. He holds a master of science degree in computer science and mathematics, and he has eight years of professional experience in software development.
Andrija is now available for hire




Belgrade, Serbia



Preferred Environment

Git, PyCharm, PyTorch, TensorFlow, Ubuntu

The most amazing...

...thing I've developed is an outfit recommendation engine which mimcks professional stylists.


  • Machine Learning Engineer

    2019 - PRESENT
    • Built next-gen HD maps for autonomous vehicles.
    • Implemented and designed state-of-the-art models for semantic segmentation, lane detection, depth estimation, etc.
    Technologies: C++, Python, TensorFlow
  • Machine Learning Engineer

    2019 - PRESENT
    Trendage (via Toptal)
    • Built a recommendation engine for the apparel industry.
    • Investigated and implemented state-of-the-art models that are able to mimic professional stylists.
    Technologies: Azure, Python, MySQL, TensorFlow, PyTorch
  • Senior Software Engineer

    2017 - 2019
    • Supported several critical back-end components that are handling more then 2 million users daily.
    Technologies: ClickHouse, Redis, Cassandra, Python, Lua, C++
  • Machine Learning Engineer

    2017 - 2017
    SparcXSoftware (via Toptal)
    • Implemented a clustering algorithm for extracting significant locations in GPS data collected from the user's phone.
    • Designed and implemented a model for clothes tagging.
    Technologies: Keras, TensorFlow, Python
  • Lead Developer

    2015 - 2017
    Future Gaming Europe
    • Helped develop a casino management system consisting of both hardware and software, with several operational modules for handling and tracking transactions and cash flow, monitoring players, and tracking staff activity.
    • Contributed to the development of the C++/Qt application that runs on Raspberry Pi, which is installed into slot machines and provides a user interface for interacting with the system.
    • Developed several back-end modules, including ticketing and a game called Jackpot.
    Technologies: PostgreSQL, Raspberry Pi, C#, Qt, C++
  • Early-stage Researcher

    2015 - 2016
    Faculty of Mathematics
    • Worked on a project called "Predicting patients' future health state," done in collaboration with several academic institutions, including the University of Geneva, University of Maribor, and Faculty of Mathematics in Belgrade.
    • Worked on development and deployment of fast, effective, and interpretable algorithms for healthcare using advanced machine learning techniques including regularized logistic regression, multi-task learning, and deep learning.
    Technologies: Python, TensorFlow, Lua, Torch, MATLAB
  • Software Developer

    2012 - 2014
    Cisco Systems
    • Worked on project Cisco Jabber. Cisco Jabber lets you access presence, instant messaging (IM), voice, video, voice messaging, desktop sharing, and conferencing.
    • Focused on the Contact search engine, a cross-platform component of Jabber.
    • Contributed to the development of CUCILync, a desktop integration that provides access to Cisco Unified Communications from Microsoft Lync. The solution extends the presence and instant messaging (IM) capabilities of Microsoft Lync by providing access to a broad set of Cisco Unified Communications capabilities, including soft-phone standards-based, video, unified messaging, audio, videoconferencing, desk-phone control, and phone.
    Technologies: C#, C++


  • Teaching Assistanceship at a Mathematical Gymnasium (Development)

    Taught algorithms and data structures to bright young mathematicians who have achieved notability in mathematics competitions.

  • Monodepth (Development)

    Unsupervised depth estimation in TensorFlow.

  • Semantic segmentation (Development)

    Semantic segmentation of images implemented in TensorFlow.

  • Newsy (Development)

    Based on the substantial amount of news articles collected from the web, and using the Bayesian naive approach, "Newsy" classifies a given article into one out of ten different categories (,, culture, life, politics, etc.).

    Technologies used include C++, Qt, Perl, MongoDB, and more.

  • Smarting (Development)

    Developed an Android app for acquiring data from a device for a startup company based in Belgrade which is developing a fully mobile, wearable device for recording and analyzing electrical brain activity and wrote drivers for a major software platform dedicated to designing, testing, and using brain-computer interfaces such as OpenVibe and BCI2000.

    Technologies used include MATLAB, Python, C++, Java, Android, and more.


  • Languages

    C++, Lua, Python, HTML5, JavaScript, HTML, Java, C, C#, A
  • Frameworks

    Qt, Boost, Unity
  • Libraries/APIs

    TensorFlow, PyTorch, React, NumPy, SciPy, Node.js, Keras, Standard Template Library (STL)
  • Paradigms

    Data Science, Agile, Kanban
  • Platforms

    AWS EC2, Amazon Web Services (AWS), Unix, Ubuntu, Raspbian, Docker, Apache Kafka, Kubernetes, Azure, Android, Raspberry Pi, Linux, Windows, Oracle, Visual Studio 2016
  • Storage

    SQLite, PostgreSQL, MySQL, ClickHouse, Cassandra, Redis, MongoDB
  • Other

    Logistic Regression, Neural Networks, Torch, Deep Learning, Machine Learning, Deep Neural Networks, Artificial Intelligence (AI), Computer Vision, Recurrent Neural Networks, Data Mining, Reinforcement Learning, Natural Language Processing (NLP), Generative Adversarial Networks (GANs), Self-driving Cars, Data Analysis, Analysis, Scraping
  • Tools

    MATLAB, PyCharm, Git, Microsoft Visual Studio, Eclipse IDE, Make, Qmake, CMake


  • Master's degree in Computer Science and Mathematics
    2012 - 2014
    Faculty of Mathematics - Belgrade
  • Bachelor's degree in Computer Science and Mathematics
    2005 - 2012
    Faculty of Mathematics - Belgrade


  • Sequence Models
  • Machine Learning

To view more profiles

Join Toptal
Share it with others