Arzuman*
Data Scientist. 1× Perplexity AI Global Hackathon winner. Building ML systems in production — fraud detection, NLP agents, and large-scale data pipelines. Budapest, Hungary.
View projectsData Scientist — Budapest, Hungary
Arzuman
Abbasov
Data scientist & AI builder.
Over the past three years I've worked at Kapital Bank — building the bank's first ML fraud detection model, an ATM route optimisation system for 1,000+ machines, and an NLP customer-service agent serving 5M+ users.
Currently pursuing an MSc in Data Science at Óbuda University, Budapest. 8 hackathon wins including 1st at the Perplexity AI Global Hackathon (USA, 2025).
Production ML at scale, from banking to global AI.
Two roles. Eight hackathon wins. One MSc in progress.
Data science in the real world.
Kapital Bank
Data Scientist · Baku, Azerbaijan
- Bank's first ML fraud detection model — 40% fewer false positives
- ATM cash route optimisation across 1,000+ machines
- NLP customer-service agent serving 5M+ users
BIRainy
Data Scientist · Baku, Azerbaijan
- Automated Power BI dashboards — 60% less manual effort
- Analysed 300M+ row datasets end-to-end
- OCR pipelines for high-volume document digitisation
Hackathons & MSc
Óbuda University · Budapest
- 1st — Perplexity AI Global Hackathon (USA, 2025)
- 1st — Teknofest Türkiye LLM App (2024)
- 1st — Heydar Aliyev 100 Engineering (2024)
- MSc Data Science — Óbuda University
Built the bank's first ML fraud detection model (–40% false positives), NLP-powered customer service agents serving 5M+ users, and LLM applications that won international hackathons.
Large-scale analysis on 300M+ row datasets. Automated financial reporting in Power BI (–60% manual effort). OCR pipelines for high-volume document digitisation.
MSc Data Science at Óbuda University. Authored a TDK research paper benchmarking open-source Azerbaijani LLMs. Strong foundation in statistical diagnostics and modelling.
Engineered an ATM cash replenishment route optimisation system for 1,000+ ATMs at Kapital Bank — combining geospatial analysis with ML demand forecasting.