HANAKOWALSKI
IT & Technology · Data Scientist

Find the
signal.

I turn noisy datasets into decisions people can trust, through statistical thinking, predictive modeling, experimentation and clear storytelling.

01 / Approach

Evidence
before ego.

Data science is a loop: frame the decision, interrogate the data, test the hypothesis, quantify uncertainty, then communicate what should happen next.

Curious about the why. Precise about the what.

My work sits where business questions meet statistical rigor. I care about useful models, honest metrics and analysis that changes an actual decision.

  1. FrameTranslate ambiguity into a measurable question and a decision criterion.
  2. ExploreFind distributions, missingness, leakage, bias and the patterns worth testing.
  3. ModelBuild interpretable baselines before reaching for complexity.
  4. ValidateStress-test assumptions, uncertainty, generalization and business relevance.
  5. ExplainMake the result legible enough for a room to act on it.
02 / Expertise

The
toolbox.

Instead of generic percentage bars, expertise is represented as rotating analytical lenses, from statistical foundations to production-minded modeling.

A / 01

Statistics & Experimentation

Hypothesis testing, causal reasoning, A/B design, confidence intervals and power-aware decision making.

A / 02

Predictive Modeling

Regression, classification, ensembles, feature engineering, calibration and robust evaluation.

A / 03

Machine Learning

Scalable pipelines, representation learning, model selection and practical model diagnostics.

A / 04

Data Storytelling

Executive narratives, exploratory analysis and visual explanations that preserve statistical nuance.

03 / Selected models

Signals
in action.

Case studies framed around the analytical question, the intervention and the decision impact, not just the library names.

CASE / 001 · RETENTION

Who is about to drift away?

Built a churn-risk model from behavioral sequences and account signals, then paired calibrated risk scores with intervention thresholds for customer teams.

PythonXGBoostSHAPCalibration
+23%

improvement in targeted retention campaign lift after replacing broad segmentation with model-ranked outreach.

Outcome / measured
CASE / 002 · FORECASTING

Forecast demand without pretending certainty.

Designed a probabilistic demand forecasting workflow with seasonality features, backtesting and prediction intervals for operations planning.

Time SeriesBacktestingPythonUncertainty
-18%

forecast error on the priority planning horizon after introducing segment-aware features and rolling validation.

Outcome / planning
CASE / 003 · EXPERIMENTS

Separate novelty from real lift.

Rebuilt an experimentation scorecard around pre-registered metrics, guardrails and sequentially aware interpretation for product decisions.

A/B TestingCausal ThinkingSQLDecision Science
36

experiments standardized into a repeatable review framework used across product and growth squads.

Outcome / process
04 / Field notes

From query
to consequence.

A career progression from analytical craft toward models and experiments that live inside real operating decisions.

Northstar Labs

Lead predictive modeling, experimentation and decision intelligence across product and commercial teams. Translate ambiguous questions into measurable analytical programs.

Vector Commerce

Built customer intelligence, demand models and experimentation foundations for high-volume commerce workflows, partnering closely with product and operations.

SignalWorks

Developed analytical datasets, dashboards and statistical models while establishing reusable analysis patterns for a growing data team.

Decisions informed74major analytical or experimental readouts.
Model lift31%best measured improvement across selected initiatives.
Experiments36standardized product experiments.
Curiosity∞questions still worth asking.
05 / Contact

Let's find
the signal.

Have a messy dataset, a prediction problem or an experiment that needs sharper reasoning? Send the context.

Good questions make good models.

I'm open to data science roles, analytical collaborations and projects where evidence needs to become a better decision.

✉ hana@example.com
☎ +48 22 555 0148
◈ Warsaw, PL