Mehmet Ali Kaya
Data Scientist · Risk Analytics & Actuarial Modelling
M.Sc. Statistics · Ph.D. Candidate, Management Information Systems (Sakarya University)
Profile
Data scientist and statistician who builds AI-driven risk analytics for insurance and finance, from raw operational data to audited, reproducible models.
Combines a B.Sc. in Insurance & Social Security (High Honors, 3rd in department, full scholarship) and an M.Sc. in Statistics (thesis on GA, PSO and ABC portfolio optimisation) with more than three years of hands-on reporting automation, data governance and KPI analysis at Bupa Acıbadem Sigorta.
Research and open-source work span extreme-value theory, regime-switching models (GMM/HMM), IFRS 17 and Solvency II measurement, reinsurance structuring, portfolio optimisation, financial machine learning and multi-agent LLM orchestration.
Works in Python and R with tested, documented pipelines: every public project ships held-out evaluation, fixed random seeds and stated limitations.
Experience
Data Management Office Specialist · Bupa Acıbadem Sigorta
Jul 2026 – Present · Türkiye · Hybrid
- Optimises data-driven workflows and builds automated analytical reporting pipelines.
- Works on enterprise data governance and data-quality controls.
- Builds quantitative operational dashboards for the business.
Corporate Efficiency Reporting Specialist · Bupa Acıbadem Sigorta
Apr 2026 – Jul 2026 · Türkiye
- Ran quantitative reporting processes and corporate performance monitoring.
- Automated recurring reports with Python and SQL.
- Built end-to-end ETL pipelines for enterprise KPI tracking.
- Applied risk-modelling techniques to business workflows to locate bottlenecks and drive process improvements.
Assistant Corporate Efficiency Reporting Specialist · Bupa Acıbadem Sigorta
Mar 2023 – Apr 2026 · Türkiye
- Analysed insurance operations, claims and premium metrics quantitatively.
- Built and maintained Excel, SQL and Python reporting workflows.
- Designed dynamic Power BI dashboards tracking medical inflation, provision duration and claim costs for several departments.
- Ran SQL root-cause analyses on provision data, found cost leakages and gave loss-ratio recommendations.
- Modelled historical claims to project future scenarios for budget and reserve planning.
- Added validation checks to keep corporate data accurate and consistent.
Academic Researcher (Data Science) · KTO Karatay University · AI Working Group
Sep 2020 – Feb 2022 · Türkiye
- Built reproducible statistical and machine-learning analyses in R for financial time series and multi-objective optimisation.
- Ran data cleaning, exploratory analysis, feature engineering and visualisation for research projects.
- Turned technical mathematical models into clear frameworks for publication and academic review together with faculty.
Student Assistant · KTO Karatay University
Apr 2019 – Jul 2019 · Konya
- Supported R programming and data-science coursework.
Intern · MERAM Sigorta Ekspertiz Hizmetleri
Jul 2018 – Aug 2018 · Konya
- Risk management and loss-adjustment work.
Intern · Social Security Institution of Türkiye (SGK)
Aug 2017 – Sep 2017 · Konya
- Human-resources and administrative data work.
Education
Ph.D. Candidate, Management Information Systems
Sakarya University · Jan 2026 – expected Jul 2030
AI-driven risk analytics and decision-support systems; machine learning and stochastic modelling in insurance technology.
Ph.D. studies, Economics (transferred to MIS)
Selçuk University · Feb 2024 – Jan 2026
Portfolio optimisation, actuarial modelling and financial machine learning.
M.Sc. Statistics · CGPA 3.22 / 4.00
Necmettin Erbakan University · 2020 – 2023
Thesis: Comparison of performance of portfolio optimization methods. Genetic Algorithm, Particle Swarm Optimization and Artificial Bee Colony on BIST 30 and S&P 500, implemented in R and MATLAB. YÖK Thesis No. 832147 · Advisor: Prof. Dr. Murat Erişoğlu.
B.Sc. Insurance & Social Security · CGPA 3.41 / 4.00
KTO Karatay University · 2015 – 2019
High Honors certificate, 3rd in department, full scholarship.
Skills
Actuarial & risk: EVT / GPD / POT / Hill · VaR, CVaR, TVaR · Solvency II SCR · IFRS 17 · copulas · Monte Carlo · GLM · frequency-severity · credit risk (PD/LGD/EAD, Vasicek) · excess-of-loss reinsurance · Lee-Carter mortality · ALM · Takaful
Statistics & ML: GMM (EM) · HMM (Baum-Welch) · regime switching · semi-Markov & Hawkes processes · time series · bootstrap & permutation inference · MCDA (TOPSIS) · gradient boosting · scikit-learn · PyTorch · CNN / U-Net / ViT / LSTM / diffusion baselines · FinBERT · A/B testing · Data Envelopment Analysis (DEA)
Optimisation: Markowitz · HRP · Black-Litterman · CVaR · GA · PSO · ABC · DE · SA · GWO · ACO · WOA · HS · CS · TLBO · JAYA · NSGA-II · multi-objective assignment scheduling
Data & MLOps: Python · R · SQL · MATLAB · VBA · Power BI · DAX · DuckDB · Parquet · medallion data lake · point-in-time and lineage controls · MLflow · pytest · Git · Streamlit · Gradio · LaTeX
AI agents & LLMs: Claude Code · Codex · DeepSeek · Gemini · Ollama local models · LiteLLM routing · multi-agent task controller with test gates, independent review and an acknowledged message bus
Projects
Biogas Actuarial Pricing Engine & Solvency II Framework · Research engine
GMM (k = 3) + HMM (k = 2) regimes and EVT/GPD tails on anaerobic-digestion telemetry (119 time points × 78 variables × 8 reactors), with IFRS 17 and Solvency II measurement. SCR of 78,372 from 100,000 Monte Carlo paths; onerous-contract detection (fulfilment cash flows 41.6 vs a baseline premium of 14.1). Public code: biogas-rspim-engine (OU-SDE calibration, HMM regimes, EVT tail risk).
QuantixActuaria · Platform
Dual-mode (Takaful and conventional) actuarial portfolio platform: five-distribution EVT, CVaR, metaheuristic optimisers and multi-LLM orchestration, covered by 104 automated tests.
Café Poliçe actuarial series (5 open repositories) · Open source
Reproducible companions to published articles: a freMTPL2 motor pricing engine (Poisson GLM vs gradient boosting, Gini 0.27 → 0.32; Gamma severity; EVT large losses), excess-of-loss reinsurance priority on Danish fire losses (Hill / POT / GPD, ξ ≈ 0.50, Monte Carlo capital, time hold-out), NOAA hail trend 1996–2025 projected to 2030, a competing-claims model and a gray-box GLM with genetic-algorithm search.
QMASE · Quantix Martial Arts Engine · Research scaffold
Licence-gated data registry and a 100-discipline catalogue; semi-Markov and Hawkes models of BJJ position transitions (Weibull dwell k ≈ 1.6); GrappleMap position-transition graph published on Hugging Face (725 positions, 1,485 transitions); UCI HAR activity baseline with subject-held-out evaluation (accuracy 0.955, macro-F1 0.954 on 9 unseen volunteers); 43 automated tests.
Multi-agent AI controller · AI engineering
Task queue that dispatches work to Claude Code, Codex, DeepSeek and a local Ollama model: per-agent git worktrees, scope ownership, independent compile/import/pytest gates, prescreen and cross-agent review, dispute protocol, acknowledged message bus with retries, a measured cost/latency/quality scheduler and a live communication dashboard. Benchmarked reviewers with planted-defect experiments.
Machine-learning baseline portfolio · Open source
Twenty-plus documented, seed-fixed baselines on public data: YOLO (COCO128), U-Net and DeepLabV3 segmentation (Penn-Fudan), Vision Transformer and CNN (CIFAR-10), DDPM diffusion (Fashion-MNIST), bidirectional LSTM (IMDb), TF-IDF + SVM spam filter, PCA, RFM + K-Means segmentation, peaks-over-threshold demand tails, Gaussian-HMM regimes and bootstrap inference.
Simba desktop companion & tooling · Product
Windows desktop companion app (private repository), a cyberpunk martial-arts theme suite (Android + Windows), a photo-to-sticker studio with a Gradio demo, Android device modules and ten small, safety-first Python CLI utilities (log redactor, password generator, dry-run file organiser and others).
Publications
- Kaya, M. A. (2026). Dolu Büyüyor, Veri Küçülüyor: Otuz Yıllık Açık Hava Verisiyle İklim Hasarını Geleceğe Taşımak [Hail is growing, data is shrinking: projecting climate losses with thirty years of open data]. Café Poliçe.
- Kaya, M. A. (2026). Zirai Don Riskinin Tahmininde Yeni Nesil Yaklaşımlar [Next-generation approaches to forecasting agricultural frost risk: stochastic processes, Markov chains and ML]. Café Poliçe.
- Kaya, M. A. (2025). Rakamların Diliyle Güven İnşa Etmek: Modern Sigortacılık [Building trust in the language of numbers: modern insurance]. Café Poliçe.
- Kaya, M. A. (2025). Sigortacılıkta 'Kutu'ların Savaşı: White, Gray ve Black Box Modeller [The battle of boxes in insurance: white-, gray- and black-box models]. Café Poliçe, print issue 64.
- Kaya, M. A. (2026). Regime-Switching Extreme-Value Pricing for Parametric Insurance on Anaerobic Digestion: An IFRS 17 and Solvency II Framework. Manuscript in preparation (target: European Actuarial Journal / ASTIN Bulletin).
Certifications (27)
- SEGEM 1st Level Actuarial Exam · score 74.82 · Oct 2018
- İTÜ Data Science Specialization · 2024
- TÜBİTAK 2209-A research proposal: machine learning in health insurance
- Business Analysis Foundations · LinkedIn · Feb 2023
- Business Analysis Foundations: Business Process Modeling · LinkedIn · Feb 2023
- Power BI Essential Training · LinkedIn · Dec 2022
- Learning Excel: Data Analysis · LinkedIn · Dec 2022
- Data Visualization in Excel · GAİB Southeastern Anatolia Exporters' Associations · Dec 2022
- Data Visualization · Global AI Hub · Aug 2022
- Python 301 · Turkcell Geleceği Yazanlar · Jun 2022
- Python 201 · Turkcell Geleceği Yazanlar · Jun 2022
- Python 101 · Turkcell Geleceği Yazanlar · Jun 2022
- Data'Speak 2021 · METU Statistics Club · Jan 2022
- Why R? Turkey 2021 · May 2021
- Actuaries Summit · Marmara University Synergy Club · Mar 2021
- Introduction to R Programming · METU Statistics Club · Feb 2021
- Communications Week #11 · IEEE İTÜ Student Branch · Jan 2021
- ROTA “Geleceğine Yön Ver” · Leaders Club · Dec 2020
- Data'Speak · METU Statistics Club · Dec 2020
- Data & Analytics Summit · BUYAK · Dec 2020
- Excel Training · Istanbul Business Institute · Mar 2019
- Effective Presentation Techniques · Istanbul Business Institute · Aug 2018
- English (B1–B2) · Istanbul Business Institute · Aug 2018
- Payroll Applications · Istanbul Business Institute · Jul 2018
- Labour Law Applications · Istanbul Business Institute · Jul 2018
- Applied Entrepreneurship Training · KOSGEB · Feb 2017
- TÜBİTAK-supported 1601 Entrepreneurship Certificate Programme (4th term) · KTO Karatay University
Honors
- High Honors certificate, KTO Karatay University
- 3rd in department, B.Sc. Insurance & Social Security
- Full undergraduate scholarship
Languages
- Turkish · native
- English · professional working proficiency
- French · beginner
Discipline
- Taekwondo · 2nd Dan black belt · Turkish Taekwondo Federation
- Muay Thai · 4th Khan
- Kickboxing · green belt
- Brazilian Jiu-Jitsu · gi and no-gi · K13BJJ & Checkmat 216
Trail ultramarathon, 50 km · Konya Half Marathon
References available on request. · Information as of November 2026.