Aleksandar Loncar, Algorithms Developer in Novi Sad, Vojvodina, Serbia
Aleksandar Loncar

Algorithms Developer in Novi Sad, Vojvodina, Serbia

Member since November 21, 2018
Aleksandar is a computer science engineer and a passionate mathematician and algorithm engineer with strong analytical and problem-solving skills. He started as a web developer and over the years found himself in the data science field, especially in deep learning.
Aleksandar is now available for hire

Portfolio

  • Usoft
    Python, C#, Tensorflow, Keras, C++, .NET, Redis
  • Brisqq LTD
    Python, Tensorflow, Keras, MongoDb
  • Danulabs
    ASP.NET, Angular, MSSQL, MySQL, Elasticsearch

Experience

  • C#, 7 years
  • Computer Vision, 3 years
  • Machine Learning, 3 years
  • Data Science, 3 years
  • TensorFlow, 3 years
  • Deep Learning, 3 years
  • Python, 3 years

Location

Novi Sad, Vojvodina, Serbia

Availability

Part-time

Preferred Environment

Windows, Linux, Git

The most amazing...

...moments I have is when I'm creating new or adjusting an algorithm for some non trivial problem.

Employment

  • Senior Software Engineer/Machine Learning Engineer

    2015 - PRESENT
    Usoft
    • Made full stack machine learning pipelines for both training and production environment in Python.
    • Involved in development of a Face Recognition system based on deep learning. A ML project developed on TensorFlow framework. Participated in training (SotA) ML models for face embeddings, blur and illuminance detection and their influence on results, occlusion detection (ex. glasses detection), fraud detection system research (liveness detection/anti-spoofing) and develop simple solution based on challenge-response protocol (smile and head pose detection), training dataset cleaning technics.
    • Developed a machine learning model based on transfer learning for leaf counting (plant phenotyping).
    • Developed many regression and classification models for tabular data.
    • Maintained, optimized and developed some features in a voice authentication system. Responsible for SIP proxy app that bridges incompatibility between the PBX and their SIP stack. Made a caching mechanism for multi machines communication based on Redis, server SIP call recording (SIP over TCP messages) for passive authentication.
    Technologies: Python, C#, Tensorflow, Keras, C++, .NET, Redis
  • Data Scientist

    2017 - 2018
    Brisqq LTD
    • Performed time series analysis and forecasting.
    • Participated in courier tiers optimization.
    • Developed model for next delivery prediction, time and geo-location. Part of the recommendation system for couriers.
    • Performed geo-location clustering based on data history.
    • Made a model that classifies couriers based on their performances.
    Technologies: Python, Tensorflow, Keras, MongoDb
  • Senior Software Developer/Development Team Leader

    2014 - 2015
    Danulabs
    • Tasked with making enterprise cloud-based software solutions for medical practices and other entities in the domain of healthcare.
    • Developed enterprise software architectures with multi-tenant database models.
    • Lead team of six people.
    Technologies: ASP.NET, Angular, MSSQL, MySQL, Elasticsearch
  • Software Developer

    2012 - 2014
    Danulabs
    • Developed custom web applications.
    Technologies: ASP.NET, C#, MSSQL, MySQL

Experience

  • Face Recognition System (Development)

    Face Recognition system is a PoC based on deep learning. Users can create profiles on the system, the number of required images for the profile can vary and depend on configuration. Next step is that the user can verify new image against the enrolled profile. The system is set to minimize false positive rate, and prevent impostors to verify as registered users. It can also execute identification against all enrolled users.

    Because it is trained on best-known datasets, it is very robust to illuminance, image blur, head pose variation, gender, and ethnicity.

  • Couriers Delivery Forecasting System (Development)

    A recommendation system for couriers based on machine learning models that can predict next delivery in spatial grid and time, based on historical data.

Skills

  • Languages

    C#, Python
  • Libraries/APIs

    Keras, TensorFlow, Pandas
  • Other

    Deep Learning, Algorithms, Mathematics, Machine Learning, Computer Vision, Algebra
  • Frameworks

    .NET
  • Tools

    Jupyter, Git
  • Paradigms

    Data Science
  • Platforms

    Windows, Linux, Docker
  • Storage

    MySQL, Redis

Education

  • Master's degree in Computer Science
    2012 - 2014
    University of Novi Sad - Novi Sad, Serbia
  • Bachelor's degree in Computer Science
    2007 - 2011
    University of Novi Sad - Novi Sad, Serbia
Certifications
  • Deep Learning Specialization
    FEBRUARY 2018 - PRESENT
    Coursera
  • Convolutional Neural Networks
    FEBRUARY 2018 - PRESENT
    Coursera
  • Sequence Models
    FEBRUARY 2018 - PRESENT
    Corsera
  • Neural Networks and Deep Learning
    AUGUST 2017 - PRESENT
    Coursera
  • Structuring Machine Learning Projects
    AUGUST 2017 - PRESENT
    Coursera
  • Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization
    AUGUST 2017 - PRESENT
    Coursera

To view more profiles

Join Toptal
I really like this profile
Share it with others