Resume

Senior Data Scientist

Publicis Sapient

Design and implement state-of-the-art solutions with focus on LLMs and structured data extraction.

• LLMs · RAG · Vector Databases · DSPy · Evaluation · Prompt Engineering · Knowledge Graphs

Senior Data Scientist

GridDynamics

Design and implement new solutions with focus on predictive analytics. Run projects in NLP domain using newest machine learning technologies.

• Data Modeling · Analytics · Artificial Neural Networks · PyTorch · Keras · TensorFlow

Advisory Board Member

Serbian AI Society

Serbian AI Society is being launched in Serbia by the most prominent people in artificial intelligence, with one goal - networking and strengthening the domestic AI scene. The aim of this organization is to bring together all stakeholders dealing with this subject which will contribute to the presence and growth of artificial intelligence in Serbia.

• AI · Growth · Networking · Strength · Serbia

AI Software Developer

Devtech

Lead AI team, mentor and drive AI efforts. Develop PoC solutions for the projects by adding AI components.

• AI · Data · Mentoring · PoC · Project Management

Machine Learning Team Lead

Selerant

Manage state-of-the-art machine learning tasks for text segmentation and classification. Shape product features according to the business requirements.

• Text classification · Semantic search · PoC · Project Management

Data Miner

Telenor

Develop different up-sell, cross-sell and churn models. Support risk models development. Accountable for segmentation, customer lifetime value calculation.

• Customer classification · Credit risk · NPS scoring

Doctor of Philosophy (Dr/Ph.D.)

Faculty of Mathematics, University of Belgrade, Serbia

Computer Science, Machine Learning, Natural Language Processing

Thesis titled: Modeling moral and emotional language aspects in conversational texts classification

• Text classification · Semantic Knowledge Bases (English, Serbian) · Conversation · Emotions · Morality

Magister of Science (Mr/Mag/M.Phil.)

Faculty of Mathematics, University of Belgrade, Serbia

Computer Science

Thesis titled: Applying classification methods on N-gram genome analysis

• Databases · Data Mining · Bioinformatics

Graduated Mathematician and Computer Scientist (B.Sc./M.Sc.)

Faculty of Mathematics, University of Belgrade, Serbia

Computer Science, Mathematics

• Programming Languages (Pascal, C, C++, Java, Prolog, Lisp) · Algorithms Architecture · Computer System Architecture · Mathematical Logic · Automata Theory · Databases · Microprocessors · Microcomputers · Assembly Language · Compilers and Interpreters · Programmable Systems Architecture · Algebra (I and II) · Analytical, Constructive and Euclidian Geometry · Calculus (I and II) · Real and Complex Functions · Numerical Methods (I and II) · Probability and Statistics

Annual Award of the Mathematical Institute of the Serbian Academy of Sciences and Arts (MI SANU) in the field of Computing for Ph.D. Students

Building an emotion lexicon for Serbian using curated language resources

Language Resources and Evaluation, Springer

SRPOL - A LEXICON BASED FRAMEWORK FOR SENTIMENT STRENGTH OF SERBIAN TEXTS

Review of the National Center for Digitization

Effective methods for email classification: Is it a business or personal email?

Computer Science and Information Systems

Agentic RAG for Technical Documents Search

Create advanced RAG pipeline for technical customer support. Project leverages usage of multiple vector DB, web search and prompt engineering techniques to produce the most relevant response to the users.

• RAG · Vector Databases · Prompt Engineering · LLMs · DSPy

Building an emotion lexicon for Serbian using curated language resources

Creating emotion lexicon for Serbian language using available language resources and manual annotation.

• Emotion lexicon · Language resources · Annotation · NLP · Research project

HR - Analytics

Extracting candidate attributes from the resume using machine learning techniques. Find similarity between candidate resume and position description. Compute demand time for opened positions.

• Named Entity Recognition · Text similarity · Transformers · BERT modeling · Regression

SRPOL

Create a word sentiment lexicon and complete framework to calculate sentiment strength for a given text writen in Serbian language.

• Word lexicon · Contextual sentiment triggers · Text segmentation · Sentiment strength

Semantic Query Parsing

Parse search queries to understand user search intention using historical data and user search preferences. Apply advanced methods for the semantic parsing to improve the relevance of the required search results.

• Query Understanding · Intention Modeling · Information Extraction · Synthetic Query Generation · Semantic Similarity

Health-Care Search & Answer

Applying advanced search algorithms to retrieve relevant search results. Experiment with different Q&A modeling techniques to provide relevant answer.

• Text similarity · Extreme classification · Q&A · Transformers · Transfer learning · BERT modeling · BioASQ

Email classification into Business and Personal categories

Classifying email messages with mixed content to Business/Personal categories.

• Text classification · BERT embeddings · Tf-Idf · Tensorflow · Research project

Hierarchical Multilabel Document Classification

Assigning textual data to categories with complex hierarchical structure and distribution.

• Document classification · Product attributes detection · Multilabel stratified sampling

Applying classification methods on N-gram genome analysis

Magister thesis in the field of bioinformatics using machine learning and data mining techniques - identification of genomic islands using the N-gram statistical behavior on the particular regions of the genome sequence.

• Magister thesis · Research project