Ph.D. Researcher — Digital Finance
Andrei-Theodor
Ginavar
Time series, tail risk, and economics of crypto markets.
About
I'm a Ph.D. student in Cybernetics and Economic Statistics at the Bucharest University of Economic Studies (ASE), and currently an Early Stage Researcher in the MSCA Industrial Doctoral Network in Digital Finance, hosted by Cardo AI in Milan.
I hold an MSc in Applied Statistics & Data Science from ASE, and I'm a researcher at the Institute of Digital Assets (IDA), a global research and education institution dedicated to digital assets and their integration into the economy and society.
My interest lies in digital and traditional finance, particularly applying quantitative methods to forecasting and risk management, particularly extreme risk. My published and ongoing work spans tail-risk forecasting, credit risk modelling, the use of LLMs and agents as research tools, and everything in between.
Publications
In the beginning was the Word: LLM-VaR and LLM-ES
Expert Systems with Applications, Vol. 295 · Elsevier · 1 Jan 2026
Recalibrating Tail Risk Forecasts under Temporal Dependence
SSRN Working Paper · 2026
Cryptocurrency Market Analysis: Insights from Metcalfe's Law and Log-Periodic Power Laws
Chapter in Handbook of Blockchain Analytics · Springer Nature · October 2025
Timeline
Positions and degrees
Early Stage Researcher
MSCA Industrial Doctoral Network — Digital Finance, Cardo AI
Milan, Italy
Benchmarking classical statistical learning against modern machine learning to forecast macroeconomic factors (interest rates, inflation, unemployment, house prices) that drive financial markets, then using those forecasts to generate what-if market scenarios.
Ph.D. Researcher
IDA Institute of Digital Assets
Bucharest, Romania
- Collecting and analysing large datasets on digital asset markets
- Developing quantitative models of price dynamics, market trends and investor behaviour
- Code development (Python, R) and verification for research projects
- Developing educational content for Quantinar and Quantlet, and maintaining IDA's websites
Ph.D., Cybernetics and Economic Statistics
Academia de Studii Economice din București (ASE)
Bucharest, Romania · Field: Statistics
Thesis: Application of Large Language Model models in the forecast of digital asset returns.
Research Assistant
IDA Institute of Digital Assets
Bucharest, Romania
- Collecting and analysing datasets on digital asset markets
- Developing quantitative models of price dynamics and investor behaviour
- Code development and verification for research projects
- Educational content for Quantinar; maintaining IDA websites
Data Analyst
GfK Consumer Panel | YouGov
Bucharest, Romania
- Preparing reports, quantitative analysis and client presentations
- Comparison and trend analysis of purchasing and consumption behaviour
- Implementing and testing new reporting solutions; automating the reporting process
- Capturing, verifying and visualising quantitative data; writing documentation
MSc, Applied Statistics and Data Science
Academia de Studii Economice din București (ASE)
Bucharest, Romania
Coursework: time series analysis, financial market modelling, data mining & big data, stochastic processes & models, visualisation techniques, econometrics. Thesis: Risk Modeling for Digital Assets.
Account Executive
Mercury360 Communications
Bucharest, Romania
- Coordinating execution of advertising campaigns and their budgets
- Centralising and analysing quantitative and qualitative field reports
- Managing client and supplier relationships; negotiating supplier rates
BSc, Marketing
Academia de Studii Economice din București (ASE)
Bucharest, Romania
Coursework: advanced marketing research, consumer behaviour, statistics, econometrics, micro & macroeconomics, finance. Thesis: Analysis of purchasing and consumption behaviour on the Romanian alcoholic beverages market, based on the SNA-Focus study.
Research
Work connected to my doctoral thesis, alongside independent projects in the same space.
AI for Scientific Discovery
A side project that grew out of using AI, including agents, to aid my own research — now developing into a full research project. The end goal: using solely open-source and open-weight models for research purposes.
Quantile risk forecasting
Tests fine-tuned time-series foundation models and LLM-based approaches against an econometric benchmark for Value-at-Risk, across a suite of backtests (traffic light, Acerbi–Szekely). Preliminary finding: AI-based models can match or surpass the true model under small samples and heavy tails — but performance is configuration-dependent.
A semi-analytic framework for DeFi long-short sizing
Models the initial sizing of a leveraged DeFi long-short position under bivariate Kou double-exponential jump-diffusion dynamics, reducing survival-path payoff moments to killed expectations solvable via a finite-dimensional Laplace-domain linear system.