Finance AI & Data Capabilities Architect · Tornio, Finland
Cloud data architect with a data-science past and a finance focus. I turn scattered spreadsheets and legacy systems into governed, automated, cost-aware platforms on Azure — then put intelligence on top that the business actually uses.
About
Boring means: the numbers are right, they arrive on time, nobody babysits the pipeline, and the finance team stopped e-mailing Excel files around. I've been a data scientist, a service manager and a cloud data architect, and today I lead finance AI & data capabilities at a global stainless-steel company.
That mix matters. I've built the models, I've run the service desk that got the calls when they broke, and I've drawn the architecture that keeps both from happening again. I design with the whole lifecycle in mind: ingestion, governance, cost, operations and the humans who will live with it.
Consultancy services
Engage me for an afternoon, a sprint or a quarter. Remote-first, hands-on, and always leaving you with something that runs.
From "we should do something with AI" to a shortlist of use-cases with real ROI — then shipping the first one. LLM agents, forecasting, anomaly detection, document automation.
Blueprint, build or rescue a modern data platform on Azure: lakehouse, pipelines, modelling, governance and CI/CD. See how I build them.
Month-end, forecasting, cost allocation, reporting — I speak both finance and engineering, and I automate the parts that are eating your controllers' evenings.
A senior pair of eyes on retainer: architecture reviews, vendor and tool selection, mentoring your engineers, and a sanity check before the big decisions.
Building data platforms
Every platform I design follows the same shape. The tools change; the shape doesn't — because it's the shape that keeps it cheap, trustworthy and AI-ready three years later.
Schemas, tests and ownership are agreed before the first pipeline runs. Surprises at month-end are a design failure, not bad luck.
Infrastructure, pipelines, models and permissions live in git and deploy through CI. If it can't be rebuilt from the repo, it doesn't exist.
Compute is sized, tagged and reported per domain. I've cut platform bills by double-digit percentages without touching a single dashboard.
Clean, governed gold tables are what make LLM agents and forecasting trustworthy. The platform is the moat; the model is the cherry.
CV
Global stainless-steel company · Tornio, Finland
Own the architecture and roadmap for finance data and AI capabilities in a global stainless-steel group: data platform on Azure, analytics and forecasting pipelines, and AI/LLM solutions embedded into finance processes. Bridge between finance leadership, data engineering and IT.
Global stainless-steel company
Designed and implemented cloud-based data solutions that streamlined analytics and automation — pipelines, lakehouse modelling, DevOps for data, and cost optimisation across the Azure estate.
Industrial & finance analytics
Predictive modelling, forecasting and analytics on industrial and finance data; took models from notebook to production and learned first-hand why platforms matter.
IT services
Ran services end-to-end — SLAs, incidents, vendors and stakeholders. The operational scar tissue behind every "boring, reliable" platform I design today.
Bachelor's degree
Microsoft
Lab
I practise what I sell: each of these was planned, built and tested through an agentic workflow — architect → coder → reviewer — with me setting direction.
One command builds a site for a Raspberry Pi, ships it to a private registry and swaps the container — reachable on the LAN and, behind a zero-trust tunnel, from anywhere.
Pull public match APIs and replay files into a per-opponent report — tendencies, threats, site splits — with sample-size guards so thin data stays quiet instead of lying.
A Claude-powered troubleshooter for ageing machines: reasons over symptoms and service history, then finds the actual parts and prices to fix it.
Browser-based drawing and playbook editor with a tiny backend and persistent storage, built by parallel agents against a shared type contract.
Contact
Tell me what's slow, expensive or untrustworthy today. Use the form — I reply with a straight opinion and, if it fits, a proposal.