Aleksandar Pantovic
Verified Expert in Engineering
Data Scientist and Developer
Belgrade, Serbia
Toptal member since July 25, 2022
Aleksandar is a highly accomplished data scientist who has successfully built and implemented machine learning algorithms, predictive models, and generative AI applications in various industries. He's passionate about machine learning and AI and their applications in real-world industry problems. Aleksandar is an expert in Python, R, SQL, PySpark, Databricks, TensorFlow, PyTorch, SAS, Power BI, Git, and others, and is experienced in the top 3 cloud vendors (AWS, GCP, and Azure)
Portfolio
Experience
- Data Science - 7 years
- Artificial Intelligence (AI) - 5 years
- R - 5 years
- Machine Learning - 5 years
- SQL - 5 years
- Python 3 - 3 years
- Docker - 2 years
- Natural Language Processing (NLP) - 1 year
Preferred Environment
Windows, Linux, MacOS, Python 3, SQL
The most amazing...
...things I've built are a fully automated, end-to-end demand forecasting pipeline and a document processing and automation service.
Work Experience
Senior Data Scientist
Independent Contractor
- Built various microservices to orchestrate the workload and perform OCR, extract information from documents (tables, checks, mentions, etc.), chunk the document, store it into a vector database, and process it through relevant controls assessments.
- Developed a fully automated demand forecasting pipeline based on statistical forecasts and machine learning models. Its workflow allows users to tweak parameters, run and compare models interactively, and analyze error metrics.
- Created a data processing pipeline in Spark to work with several million data points of hourly data.
- Optimized multiple R scripts that do weekly processing, thus cutting the processing time by half, and created the distribution planning logic to optimize warehouse inventory levels and re-route existing inventory if possible.
- Built a claim processing flow by utilizing an LLM to analyze contracts and business logic to determine rules for estimating allowable payments.
Data Scientist
Treat Technologies, Inc
- Developed a classification model for each client to determine whether the client belongs to a high, low, or mid-value bucket.
- Managed customer lifetime value estimation and estimated expenditures for the next 12 months.
- Implemented all models and documented code and process.
Senior Risk Modeling Associate
Banca Intesa Serbia
- Developed predictive models for the probability of default (PD) and exposure at default (EAD) estimation for retail and corporate clients utilizing logistic regression, decision trees, and ensemble models. Implemented them into IT infrastructure.
- Calculated IFRS 9 migration matrices, estimated TTC PDs, and automated the whole process flow, migrating it from complex Excel sheets to R.
- Monitored models' performance monthly, reported to internal and external stakeholders, and optimized processes by automating rating calculations and data quality controls. Migrated the entire process from Excel workbooks to Power BI.
- Participated in recruitment and training of new interns and supervised projects assigned to them.
Risk Data Scientist
Addiko Bank AG
- Built predictive models for loan approvals and capital allocation, implemented these models, and performed monthly performance monitoring. Monitored the performance of retail credit portfolios of all AG SEE banks.
- Created a completely automated model performance report for all member banks of AG SEE banks.
- Developed a predictive model for the HR department to help them in their selection process.
Reporting Anlayst
Nielsen
- Learned to manipulate large datasets and derive insights from data that gave the sales team a better negotiating position and the ability to seal better deals for the company.
- Automated most reporting processes, cutting report generation time by 40 to 50%.
- Transitioned old reports into new templates that were easier to maintain and looked more professional.
Experience
Automated Scorecard Builder
Real Estate Appraisal App
required data, the model estimates the price for an apartment and also makes comparisons with average values for that municipality. Also, a geolocation service is included in the app. The data was scraped from local real estate listing websites.
Time Series Forecasting in Spark
https://github.com/apantovic/spark_tsPipeline can be also adjusted to accommodate different types of regression/classification problems.
Education
Master's Degree in Statistics
University of Belgrade - Belgrade, Serbia
Skills
Libraries/APIs
Scikit-learn, Pandas, TensorFlow, Keras, XGBoost, Matplotlib, LSTM, OpenAI API, PySpark, Dask, REST APIs, PyTorch, Three.js
Tools
Microsoft Power BI, Seaborn, BigQuery, Apache Airflow, Claude Code, Azure Machine Learning, Excel 2013, Looker, Amazon SageMaker, Tableau
Languages
R, SQL, Python 3, SAS, Python, Excel VBA
Platforms
Databricks, Jupyter Notebook, Azure, Docker, Amazon Web Services (AWS), Google Cloud Platform (GCP)
Storage
JSON, MySQL, PostgreSQL, Data Pipelines
Frameworks
Apache Spark, Spark, LangGraph
Paradigms
ETL
Industry Expertise
Project Management
Other
Time Series, Time Series Analysis, Forecasting, Machine Learning, Data Science, Predictive Modeling, Predictive Analytics, Data Analysis, Data Visualization, Data Analytics, Data Reporting, Data Mining, Big Data, Statistics, Quantitative Finance, Finance, Random Forest Regression, Linear Regression, Random Forests, Supply Chain, Supply Chain Management (SCM), Supply Chain Optimization, Exploratory Data Analysis, Financial Data, eCommerce, Statistical Analysis, Statistical Modeling, Dashboards, Excel 365, Agentic RAG Systems, Large Language Models (LLMs), Feature Engineering, Statistical Methods, APIs, RAG Systems, Pgvector, Inventory Management, Quantitative Analysis, Reports, Artificial Intelligence (AI), Neural Networks, Natural Language Processing (NLP), Web Scraping, Data Engineering, Google BigQuery, Financial Modeling, Machine Learning Operations (MLOps), Risk Analysis, Generative Pre-trained Transformers (GPT), Recommendation Systems, Applied Mathematics, Pricing, Reporting, LangChain, FastAPI, Agentic AI, Anthropic, Prompt Engineering, Retrieval-augmented Generation (RAG), Vector Databases, AI Agents, API Integration, Decision Modeling, Supabase, AI Agent Orchestration, AI Architecture, Optical Character Recognition (OCR), User Behavior, Gradient Boosting, Data Build Tool (dbt), Vehicle Routing, Geographic Information Systems
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring