Vladisav Jelisavcic, Developer in Belgrade, Serbia
Vladisav is available for hire
Hire Vladisav

Vladisav Jelisavcic

Verified Expert  in Engineering

Machine Learning Developer

Location
Belgrade, Serbia
Toptal Member Since
January 10, 2017

Vladisav received his PhD in computer science from Belgrade University—specializing in machine learning with big data. He is passionate about big data and scalable machine learning algorithms and possesses expertise in Python and Java. He has over a decade of experience in programming and machine learning while working in teams of varying sizes. Vladisav enjoys working in innovative and fast-paced startup/academic environments.

Portfolio

Censia
Amazon Web Services (AWS), Big Data, Machine Learning, Data Analytics, Pandas...
Mathematical Institute of the Serbian Academy of Sciences and Arts
Big Data, Machine Learning, Data Analytics, Linux, Pandas, Python, Apache Spark...
Magnaf.ai
Amazon Web Services (AWS), Big Data, Machine Learning, Data Analytics, Linux...

Experience

Availability

Part-time

Preferred Environment

Jupyter, Git, Apache Maven, IntelliJ IDEA

The most amazing...

...algorithm I've created and coded is a novel tool for learning the scale-free Gaussian Markov random fields from big data.

Work Experience

Data Scientist

2017 - PRESENT
Censia
  • Derived meaningful insights from semi-structured big data in the terabyte range.
  • Implemented NLP algorithms for semantic clustering, conflation, and text understanding.
  • Designed and scaled-up novel algorithms for scalable machine learning analytics on terabytes of data.
  • Visualized insights and patterns.
  • Derived and delivered product features from data.
Technologies: Amazon Web Services (AWS), Big Data, Machine Learning, Data Analytics, Pandas, Python, Apache Spark, Jupyter, TensorFlow, Keras, Amazon S3 (AWS S3), Spark ML, Git, Data Mining, Research, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), GPT, REST, Scikit-learn, Jupyter Notebook, Databricks, Plotly, Spark, PySpark, PyTorch, NumPy, Data Engineering, Zeppelin

Research Assistant

2012 - PRESENT
Mathematical Institute of the Serbian Academy of Sciences and Arts
  • Developed a software tool for early aneurysm prediction using neural networks and Apache Spark.
  • Worked on a prediction model for analyzing tensiomyography signals.
  • Worked as the lead back-end developer for a wiki-based system for a cultural heritage presentation with rich metadata management. The system was written in Java using eXistdb for storage.
  • Implemented a module for topic modeling.
  • Developed a software framework in Java for complex dynamic form generation based on XSD schema.
  • Worked on accelerating and improving scalability of machine learning algorithms for L1 regularized structured prediction.
  • Ran the research on novel word embedding methods for various NLP tasks.
Technologies: Big Data, Machine Learning, Data Analytics, Linux, Pandas, Python, Apache Spark, Jupyter, Apache Ignite, TensorFlow, Keras, Amazon DynamoDB, Spark ML, JavaScript, C, Git, Apache Tomcat, Data Mining, Research, Generative Pre-trained Transformers (GPT), GPT, Natural Language Processing (NLP), Object-oriented Programming (OOP), REST, Scikit-learn, Jupyter Notebook, Concurrent Programming, Distributed Programming, Databricks, MALLET, Mahout, Plotly, Spark, PySpark, PyTorch, eXist-db, Octave, MATLAB, Apache Lucene, Spring, Hibernate, JPA, JSF, Jakarta Server Pages (JSP), Java

Lead Technical Engineer

2016 - 2018
Magnaf.ai
  • Designed and implemented scalable microservice platform for image processing and entity recognition from the text (Python/Node.js, Google Cloud, Docker, TensorFlow, Keras, and Scikit-learn).
  • Designed a data processing platform with the semantic matching of extracted entities, based on CQRS and event-sourcing paradigm (Python/TensorFlow, and Elasticsearch).
  • Created and optimized REST API using Node.js and Swagger for several bottleneck services.
  • Defined data collection, preprocessing, and annotation workflows, and outlined the design for annotation tool custom made for the domain.
  • Created deployment scripts and procedures (Docker, Node.js, Google Cloud App Engine, AWS Beanstalk, and ZEIT).
Technologies: Amazon Web Services (AWS), Big Data, Machine Learning, Data Analytics, Linux, Pandas, Python, Jupyter, Google Cloud Storage, JavaScript, AWS Lambda, Git, Data Mining, Research, GPT, Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Object-oriented Programming (OOP), REST, Scikit-learn, Asynchronous Programming, Google Cloud Datastore, PubSubJS, Google Cloud Functions, Event Sourcing, CQRS, Docker, Google Cloud, Serverless, Elasticsearch, Node.js

Software Developer

2012 - 2016
Vitid
  • Implemented an ERP system in Java.
  • Worked on a business intelligence module.
  • Developed a human resources and employee payroll module.
Technologies: Data Analytics, Linux, Git, Apache Tomcat, Object-oriented Programming (OOP), PostgreSQL, MySQL, Hibernate, JPA, JSF, Jakarta Server Pages (JSP), Spring, Java

Research Scholar

2014 - 2014
Center for Data Analytics and Biomedical Informatics | Temple University
  • Developed an algorithm for fast learning of a sparse Gaussian Markov Random field from big data.
  • Worked on developing novel algorithms for fast structured regression on large heterogeneous graphs.
  • Built a large-scale prediction model for sepsis prediction from inpatient hospital data.
Technologies: Big Data, Machine Learning, Data Analytics, Pandas, Jupyter, Data Mining, Research, Scikit-learn, Jupyter Notebook, NumPy, Data Engineering, Octave, MATLAB

Apache Ignite

https://ignite.apache.org/
An Apache committer for the Apache Ignite project—one of leading distributed in-memory platforms for computing and transacting on large-scale data sets in real-time.

Decomposition-based Reparameterization for Efficient Estimation of Sparse Gaussian Conditional Random Fields

https://bit.ly/2ZuaAlx
Simultaneously estimating a multi-output regression model—while recovering dependency structure among variables from high-dimensional observations—is an interesting and useful exercise in contemporary statistical learning apps. A prominent approach is to fit a Sparse Gaussian Conditional Random Field by optimizing regularized maximum likelihood objective, where the sparsity is induced by imposing L1 norm on the entries of a precision and transformation matrix.

We studied how the reparametrization of the original problem may lead to more efficient estimation procedures. Particularly, instead of representing a problem through a precision matrix, we used its Cholesky factor—its attractive properties allowed inexpensive coordinate descent-based optimization algorithm that is also highly parallelizable.

Learning of Scale-free Networks based on Cholesky Factorization

https://bit.ly/3bzfuW1
Recovering network connectivity structure from high dimensional observations is becoming more vital in statistical learning apps. A prominent approach is to learn a Sparse Gaussian Markov Random Field by optimizing regularized maximum likelihood, where the sparsity is induced by imposing the 𝐿1 norm on the entries of a precision matrix.

In this article, we shed light on an alternative objective, where instead of precision, its Cholesky factor is penalized by the 𝐿1 norm. We show that such an objective criterion possesses attractive properties that allowed us to develop a very fast scale-free network estimation via a Cholesky factorization (SNETCH) optimization algorithm based on coordinate descent (which is highly parallelizable and can exploit an active set approach). This is particularly suited for problems with structures that allow sparse Cholesky factor, crucial for scale-free networks.

Evaluation of synthetically generated examples and high-impact apps from a biomedical domain of 900,000+ variables provide evidence that for such tasks the SNETCH algorithm can learn the underlying structure more accurately, and an order of magnitude faster than state-of-the-art approaches based on the 𝐿1 penalized precision matrix.

Fast Sparse Gaussian Markov Random Fields Learning Based on Cholesky Factorization

https://bit.ly/2ZfKaqy
Learning the sparse Gaussian Markov Random Field, or conversely, estimating the sparse inverse covariance matrix is an approach to uncover the underlying dependency structure in data. Most of the current methods solve the problem by optimizing the maximum likelihood objective with a Laplace prior L1 on entries of a precision matrix.

We proposed a novel objective with a regularization term that penalizes an approximate product of the Cholesky decomposed precision matrix. This new reparametrization of the penalty term allows efficient coordinate descent optimization. It results in synergy with an active set approach that results in a fast and efficient method for learning the sparse inverse covariance matrix.

We evaluated the speed and solution quality of the newly proposed SCHL method on problems consisting of up to 24,840 variables. Our approach was several times faster than three state-of-the-art approaches. Applying it to a high impact problem from the health informatics domain, we also demonstrate that SCHL can be used to discover interpretable networks.

Distance-based Modeling of Interactions in Structured Regressions

https://bit.ly/3h6JSID
Graphical models—as applied to multi-target prediction problems—commonly utilize interaction terms to impose structure among the output variables. Often, such construction is based on the assumption that related outputs need to be similar and that we adopt interaction terms that force them to be closer.

Here, we relax that assumption and propose a feature that is based on distance and can adapt to ensure that variables have a smaller or larger difference in values. We utilized a Gaussian conditional random field model, where we have extended its originally proposed interaction potential to include a distance term. The extended model is compared to the baseline in various structured regression setups. An increase in predictive accuracy was observed on both synthetic examples and real-world applications, including challenging tasks from climate and healthcare domains.

SNETCH

https://github.com/vladisav/SNETCH
I wrote this code for an algorithm to recover the network connectivity structure from high-dimensional scale-free observations.

Languages

Python, Java, Octave, C, JavaScript

Frameworks

Spark, Apache Spark, Spring, Hibernate, JPA, Jakarta Server Pages (JSP), JSF

Libraries/APIs

PySpark, NumPy, PyTorch, Pandas, Scikit-learn, TensorFlow, Keras, Spark ML, PubSubJS, MALLET, Mahout, Apache Lucene, Node.js

Tools

Git, Apache Ignite, MATLAB, Jupyter, Plotly, IntelliJ IDEA, Apache Maven, Apache Tomcat

Paradigms

REST, Concurrent Programming, Object-oriented Programming (OOP), Distributed Programming, Asynchronous Programming, Event Sourcing, CQRS

Platforms

Linux, Databricks, Zeppelin, Jupyter Notebook, Docker, AWS Lambda, Amazon Web Services (AWS)

Storage

Elasticsearch, Google Cloud Datastore, Google Cloud Storage, Amazon DynamoDB, Amazon S3 (AWS S3), MySQL, eXist-db, Google Cloud, PostgreSQL

Other

Data Engineering, Machine Learning, Data Analytics, Data Mining, Natural Language Processing (NLP), Big Data, Research, GPT, Generative Pre-trained Transformers (GPT), Serverless, Google Cloud Functions

2012 - 2018

PhD Degree in Electrical Engineering and Computer Science

University of Belgrade, School of Electrical Engineering - Belgrade, Serbia

2010 - 2011

Master's Degree in Electrical Engineering and Computer Science

University of Belgrade, School of Electrical Engineering - Belgrade, Serbia

2006 - 2010

Bachelor's Degree in Electrical Engineering and Computer Science

University of Belgrade, School of Electrical Engineering - Belgrade, Serbia

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring