Firaz Zakariya
Data scientist and analyst based in Karlstad, Sweden. I build forecasting, classification, and experimentation systems where the point is a decision: capacity plans, risk prioritisation, rollout calls, not a leaderboard score. Most of my work sits at the intersection of statistical rigour and interpretable models, from hybrid time-series forecasts to SHAP-backed risk classifiers to properly powered A/B tests.
Experimentation, forecasting, and analytics engineering. Start with microexp →
Projects
microexp
Sequential and Bayesian A/B testing for low-traffic products
A Python package for running rigorous A/B tests when you have hundreds of users per week, not millions. Sequential testing (mSPRT), Bayesian methods, CUPED variance reduction. Born from running experiments on Cawosh, my SaaS side project.
Battery degradation modelling for EVs
Four-model comparison with prognostic evaluation
A reproducible comparison of four model families (statistical baseline, gradient-boosted, LSTM, and physics-informed) for predicting lithium-ion capacity fade under real EV driving profiles. Evaluated with prognostic metrics and cross-dataset generalisation.