Milan Zarchev
Data scientist | PhD candidate
↓
About me
Hi, welcome to my portfolio page. My name is Milan and I specialize in converting data to decisions. I have been juggling data for over 8 years, using cutting edge methods to deliver robust results to technical audiences, academic journals and business stakeholders.
I am an expert in R and can fluently handle tabular data (Excel, SPSS) or large relational SQL data. I have a deep understanding of data science and machine learning through their full life-cycle and the business impact they can have. Although I take pride in my technical prowess, I firmly believe it alone cannot overcome critical thinking skills applied to the context behind the data and articulate communication tailored for the audience. In my work I strive towards a high degree of autonomy while using data responsibly and in accordance with each company's values.
Bachelor's (Honours)
Psychology
Master's (Cum laude)
Research in Health Sciences
PhD candidate
Computational psychiatry
Visual storytelling
My favorite part of any project is using data to tell stories by highlighting what is relevant to my audience and inviting actionable discussion
Statistical modeling
I have rich professional experience both with traditional models for explaining patterns or machine learning algorithms for predicting trends
Dashboard making
Some projects are best presented in an interactive dashboard. In those cases, I use Shiny to present and deploy results, leaving you to intuitively interact with your data
Technical skills
The process of translating data into actionable insights has a big component of technical proficiency to it. In this section I list the frameworks I make use of for the primary steps in a data project. Read on if you are curious what goes on behind the scenes:
My primary languages are R and SQL for manipulation, cleaning and preparation of data. For traditional statistical modeling, I use STAN Bayesian models or regular Frequentist models if a p-value is needed. For supervised and unsupervised machine learning I use caret and the tidymodels framework. I rely on ggplot2 for visualization and shiny in combination with flexdashboard for custom dashboards. PDF reports, Word documents, Powerpoint presentations and other reports are created within Markdown for reproducible results that can effortlessly incorporate parameter changes.
