Pavel L.Private profile: Upwork shows its earnings only to signed-in members, so we check it every few weeks, not daily. MRR here is a rough estimate (Upwork's six-month earnings ÷ months active). Last read Nov 5, 2026.

I solve quantitative problems in forecasting, pricing, risk, and causal inference — including the cases without off-the-shelf solutions: demand forecasting with censored sales data, price elasticity from observational data, risk models that need to be defensible to regulators, prediction systems where a standard approach fails and someone needs to figure out why. My toolkit is statistical and probabilistic modeling, time series methods, and causal inference, built over a career that runs from professional software engineering through academic research into applied data science. That range means I can deliver whatever a project needs, from a production pipeline to a board-ready report. I pick the method that answers the question, not the one I prefer, and I go deep on the domain the problem lives in. In practice, that has meant four years modeling demand and price elasticity across hundreds of thousands of SKU-site combinations at a UK pricing consultancy; two Shopify apps (Sell Smart for price optimization, Inventory Guru for demand forecasting); and prior to that, modeling EEG and eye-tracking data in academic research. The common thread is hard inference problems where the data is messy, the standard model doesn't fit, and someone has to think carefully about what's actually being estimated. Most engagements fall into one of these areas: - Demand forecasting (including censored/stockout-corrected data) - Price elasticity estimation and pricing strategy - Causal inference and incrementality testing - Marketing mix modeling / media attribution - Risk scoring and classification models - Time series modeling and anomaly detection - Statistical consulting for academic research - Production ML pipelines (Python, R) The work usually starts with a short call to understand the problem, the data, and what would count as a good answer, followed by a written scope doc covering the proposed method, what I'd need from you, the deliverable, and where the method might fall short. From there it's iterative, with regular check-ins so you see intermediate results and can redirect. The final handoff is a Python or R pipeline that runs on your data, a backtest showing performance on held-out data, and a short document covering the model, its assumptions, and where it breaks — plus a walkthrough so you can run and interpret it on your own. PhD in Computational Cognitive Science. 12+ years of applied modeling. 51 Upwork jobs, $100K+ earned. NumPyro | JAX | Stan | Python | R | Shopify | SQL | Bayesian | Time Series | Causal Inference