Independent data practice

Scattered data,
clear decisions.
Built to ship.

I'm a freelance data scientist and engineer. I take the messy data your business already has — and turn it into models, pipelines and dashboards your team can actually use.

Start a project → See what I do

Science · models & forecasting

Engineering · pipelines & infra

Analytics · metrics & reporting

What I do

Three roles, one engagement.

Most data problems don't fit neatly into one job title. I cover the full stack — from raw collection to the decision it informs — so you don't have to coordinate three contractors.

01 — Data Science

Models that answer questions

Predict, classify, segment, forecast. Built to be measured against a real baseline — not just demoed.

Forecasting & demand planning
Churn, risk & propensity models
Customer & product segmentation
Experiment design & A/B analysis

02 — Data Engineering

Pipelines that don't break

The plumbing that gets clean, trustworthy data to the right place on schedule — and tells you when something's wrong.

ETL / ELT & data warehousing
API & database integrations
Cloud deployment & automation
Data quality & monitoring

03 — Analytics

Numbers people act on

Dashboards and analysis that make the next decision obvious — defined around the metrics that actually move your business.

KPI & reporting dashboards
Exploratory & ad-hoc analysis
Metric definition & tracking
Self-serve reporting for your team

Selected work

From messy export to decisions.

An end-to-end case study on real transaction data: the same process I'd run on yours, from the first look at the raw files to a dashboard the team uses.

Case study · Retail & e-commerce

Retail Intelligence

A UK online wholesaler had two years of sales in a spreadsheet: 1,067,371 messy invoice lines. I built a tested pipeline into a DuckDB warehouse, checked the owner's assumptions statistically, modelled customer value and purchase propensity, and delivered an interactive dashboard.

+2.2%growth, not significant: the business is flat

69%of revenue from 22% of customers

2.1×more buyers reached by calling the model's top picks

PythonSQLDuckDBscikit-learnstatsmodelsStreamlit

Try the live dashboard → Read the case study View the code

Retail Intelligence dashboard showing revenue, orders and customer KPIs above a weekly revenue chart

Case study · ERP & finance operations

Procure-to-Pay Intelligence

Real ERP records from a multinational manufacturer: 1.6 million events on a €997M order book. I mapped what actually happens between ordering and paying, tested the financial controls on every order line, and used causal and predictive models to show what really drives delay and when cash goes out.

€200Mof orders with a control exception to review

5 daysapparent rework delay that vanished once adjusted

3.6%average miss on 30-day cash-out forecasts

dbtProcess miningCausal MLBayesian modelsPyTorchStreamlit

Try the live dashboard → Read the case study View the code

Procure-to-Pay Intelligence dashboard showing a ranked work queue of open order lines with the reason each is flagged

How it works

A clear path, every time.

No surprises and no lock-in. Each stage has a defined output you can review before we move on.

STEP 01

Scope

A short paid discovery, or a free intro call. We agree the question, the data available, and what a useful result looks like — in writing.

STEP 02

Build

I work in focused increments with regular check-ins. You see progress early; scope stays flexible without becoming a moving target.

STEP 03

Validate

Every model or pipeline is tested against a baseline and documented. We confirm it does the job before it goes anywhere near production.

STEP 04

Hand over

You get clean code, documentation, and a walkthrough — so the work keeps running whether I stay involved or not.

Toolkit

Tools I reach for.

Chosen to fit the problem and your existing setup — not the other way around.

Python SQL pandas / Polars scikit-learn PyTorch dbt Airflow Snowflake / BigQuery Postgres AWS / GCP Docker Power BI / Looker Streamlit Git

About

Hi, I'm Cloud.

I spent years working with data inside teams — and saw how often good data work stalls between the people who have the data and the people who need the answer.

Now I work independently with small and mid-sized businesses that need serious data capability without hiring a full team. You get one accountable person who can model the problem, build the pipeline, and explain the result in plain language.

I care about work that ships and keeps working after I leave. No black boxes, no jargon for its own sake — just clear, durable results.

Based inEuropean Union · remote-first

BackgroundData Science & IT

EngagementsProjects · Retainer · Advisory · Consulting

AvailabilityTaking new projects

LanguagesEnglish

Get in touch

Have data sitting unused?

Tell me what you're trying to figure out. I'll reply within two business days with whether — and how — I can help.

› stratumfreelance@gmail.com