Data science consulting
that ships to production.
We work on the full stack — from problem framing and data strategy to model development and deployment. Open-source tools only. Your code, your models, your infrastructure.
# how every engagement works $ git clone your_problem ✔ framing done # what's the actual business question? ✔ data audit # what do you have, what's missing? ✔ model built # python / R / open LLM, reproducible ✔ deployed # fastapi · docker · gitlab CI · on-premise ✔ handed over # docs + training, team is autonomous $ echo "result: yours to own and extend"
our process
Four phases, one objective: production
Every engagement follows the same discipline — from the first conversation to the last line of documentation.
Strategy & framing
We start by identifying the real business problem — not the data problem someone thinks they have. What's the decision this model needs to support? What does success look like in production? What data exists, and what's missing?
Analysis & modelling
Exploratory analysis, feature engineering, model selection and validation. We use Python (scikit-learn, PyTorch, statsmodels) or R (tidymodels, brms) depending on what fits your problem. Experiments tracked with MLflow from day one.
Development & deployment
Production-grade code with uv-managed environments, GitLab CI/CD pipelines, containerised with Docker, exposed via FastAPI or served via Streamlit/Dash. On-premise or your cloud — your call. No SaaS dependency.
Transfer & autonomy
Documentation, code review sessions and hands-on training so your team can maintain, extend and re-train the system. The engagement ends when your team doesn't need us — and comes back for the next challenge.
services
What we build
Every service uses open-source tooling, ships with reproducible environments and is handed over with documentation.
uv for dependency management, MLflow for experiment tracking, FastAPI for serving, Docker for containerisation, GitLab CI for automated pipelines.open source stack
Everything we use is open source
No black boxes. Every tool is inspectable, forkable, and replaceable. You're never locked in.
philosophy
Why open source only
A constraint that protects you — not us.
uv lockfiles for Python, renv for R, Docker for deployment. Your GitLab pipeline runs the same code in dev, staging and production — no drift.track record
200+ projects delivered since 2012
Energy, finance, healthcare, retail, research, telecoms — production systems running at scale.
Tell us about your project.
A few lines about your data, your problem and your constraints — we respond within 48 hours with an honest assessment of what's feasible and what it takes.