Quantitative Models. Better Decisions.

Delivered Bayesian forecasting across Finnish air traffic, commercial real-estate customer projections and macroeconomic projections for a web3 company — producing full predictive distributions rather than point estimates, so teams can plan against the full range of uncertainty. Built with PyMC, Stan and Pomegranate across Bayesian trees, Bayesian networks and Bayesian time-series models, and a heavy user of XGBoost and PyTorch.

Took loan pricing optimization end to end for Finland's largest bank — from research data science to production code and day-to-day operations — measuring the impact of changes with econometric pseudo-experiments.

Built a large-scale, dynamic probabilistic system that prices the BTC options market accurately, covering the vast majority of options traded and re-valuing them every hour. By comparing fair value against the market, it surfaces mispricings as they appear.

Priced Finnish air-traffic landing charges as a corporate-finance problem for a regulated monopoly — building the models and analysis that set defensible charges within the regulatory framework.

Built an automated loan-decision risk engine for collateralized consumer lending at Finland's largest bank. As in-house head of risk for a web3 company, owned both the research and the day-to-day operations of hedging. Managed multiple portfolios with Bayesian risk management, portfolio optimization and stochastic-volatility models.

Designed and ran the day-to-day operations of an ongoing pricing treatment-effect research program for Finland's largest bank, estimating customer purchase probability through econometric pseudo-experiments — difference-in-differences, regression discontinuity and instrumental variables. Building causal models that separate correlation from cause turns pricing, marketing and policy decisions into choices backed by defensible evidence.

Trained and fine-tuned several local language models, and built a retrieval-augmented generation (RAG) system grounded in company-specific data — so the model answers from an organization's own knowledge rather than generic web text, with everything running on infrastructure you control.

Deep experience processing big data through parallelization, delivered across multiple companies. If a workload can be crunched locally, I know how to parallelize it to get there; when local compute just won't cut it, I scale out with PySpark and Databricks — taking raw sources all the way to analytics-ready, well-tested datasets.

Designed numerous databases and built data pipelines for clients spanning publicly listed companies to Finnish Mittelstand, giving each a clean, dependable foundation for analytics and operations.

Senior Python developer with production cases across multiple companies, shipping services and ML workloads that run reliably — not just in notebooks. As consulting CTO for an energy startup, set up the full Linux stack end to end: laptops, physical servers and networks.

Involved in multiple SAP, Dynamics CRM integration and analytics projects — the largest a multi-year engagement for a Finnish mining-technology company — plus Odoo administration for an energy startup. The throughline: getting model output and data to flow into day-to-day operations instead of sitting in a report.

Was the market maker for EUROe, a Finnish stablecoin — operating a liquidity portfolio across multiple DeFi markets and chains, managing inventory, spreads and on-chain execution under real capital risk.

Built numerous analyses of on-chain transactions, market prices and volume, turning raw chain data into actionable insight — and helped design crypto tokenomics, shaping the incentives and mechanisms behind the protocols themselves.