Profileimage by Nikola NikodijevicMirkov Data scientist and product manager from Thalwil

Nikola Nikodijevic Mirkov


Last update: 20.02.2023

Data scientist and product manager

Company: Autonomous Softwares GmbH
Graduation: PhD in Economics and Finance
Hourly-/Daily rates: show
Languages: German (Native or Bilingual) | English (Native or Bilingual) | Italian (Native or Bilingual) | Serbian (Native or Bilingual)


HTML JavaScript (Programming Language) Agile Methodology Amazon Elastic Compute Cloud Cascading Style Sheets (CSS) DevOps Python (Programming Language) PostgreSQL MySQL NumPy + 16 more keywords




Agile, EC2, backend, CSS, Celery, DevOps, Docker, FastAPI, Flask, HTML, Javascript, MySQL, NumPy, Pandas, Plotly, Postgres, Python, Pytorch, SQL, SQLAlchemy, SciPy, Scikit-Learn, Scrum, Selenium, TensorFlow, Wireframing

Project history

02/2022 - Present
Led the tech team that designed, launched and iterated on the cross-platform app that employees can use to access their earned wage.
AKU (Banks and financial services, < 10 employees)

Designed, coded and maintained microservices that talk to the cross-platform app, charge users' debit cards, connect to Mexican payment system, automate tasks, and execute bots. I was in charge of coding the backend in Python and deploying it as a FastaAPI app in a Docker container on an EC2 instance on AWS. I also implemented CICD using webhooks to pull the code from Git automatically on every merge.

05/2019 - 01/2022
Developed and trained a predictive classification model for sentences in macroeconomic research reports.
Swiss National Bank (Banks and financial services, 1000-5000 employees)

Coded the internal web app (HTML + CSS + Javascript) that would parse a PDF of a macro research report, extract all the sentences, and assign to each one of them a probability of being "relevant". Deployed on an Ubuntu server using NGINX as a reverse proxy. I built the model using a pre-trained BERT for encoding the sentences and a Vector Support Classifier for classification. The entire backend was written in Python.

03/2015 - 01/2022
Managed a model suite for forecasting short-term GDP growth in major world economies.
Swiss National Bank (Banks and financial services, 500-1000 employees)

Some of the most relevant models I developed and enhanced were Dynamic Factor Model, Mixed-Frequency Sampling model, a medium-size Bayesian VAR, and machine learning techniques for trees (most notably, XGBoost). Building and training these models made me learn a ton about data analysis. The data ranged from hard economic data (e.g. Industrial Production) to surveys (e.g. Purchase Manager Index) to financial data (e.g. stock prices) and big data (e.g. satellite images). The codebase was written in Python, MATLAB, and Eviews.

Local Availability

Only available for remote work
Profileimage by Nikola NikodijevicMirkov Data scientist and product manager from Thalwil Data scientist and product manager