About me

Career timeline

• 2024–present

  Payroll & scheduling automation — freelance (DD84)

  Python automation for Icelandic payroll and scheduling; AL ↔ Azure Functions ↔ Business Central bridges; ETL into Azure SQL with exports for Business Central, DK, H3.

• 2024

  Data Scientist — visitor scheduling optimization (public-benefit organization)

  Built a Python/pandas optimizer placing ~500 people across 8 hotels and bus tours; real-time Keflavík flight-arrival monitor (Docker/AWS/Terraform); cost-optimized airport transfers.

• 2021–2024

  Senior Manufacturing Technician — Alvotech

  Promoted to Senior within 12 months; qualified GMP Trainer; led deviation investigations (root-cause analysis, CAPA); completed data-security and data-integrity certifications in a regulated (GMP) environment.

• 2016–present

  Accountant (volunteer) — public-benefit organization

  Full bookkeeping, ongoing.

• 2013–2021

  Relocated to Iceland; intensive self-directed learning

  Intensive self-directed learning of Python and data analysis applied to real problems; set up and advised on a small business's Icelandic tax filing (RSK); completed NTV accounting course.

• to 2013 (Poland)

  Applied-maths foundation & early career

  MSc thesis on Black–Scholes martingales; bank-systems programmer (final year of studies) and mortgage-department internship; ran own business (property insurance for PZU + web/e-commerce development), self-settling VAT including mixed-activity proportions, corrections, and fixed-asset depreciation.

The Integration Bridge

On the Icelandic market, systems like Launakerfi DK, H3, and Microsoft Business Central dominate. They're capable, but rigid — they often require manual entry of shifts, overtime, and sick days, forcing accountants into extra programs and spreadsheets. Instead of pushing you into a costly system change, I build a non-invasive bridge: Python that pulls data from wherever it lives (Excel, Azure SQL), processes it against your specific union agreements, and produces files ready to import straight into DK or H3. For Business Central, I write AL extensions that delegate heavy computation to Azure Python Functions and return a clean result to your ERP. Whatever you already run, I connect to it — fast, AI-assisted, at a fraction of the usual cost.

How I think

I've worked with data from pharmaceutical manufacturing, logistics, finance, and payroll. That range taught me to read a business's real needs quickly and to see connections between pieces of information that others miss. What sets me apart isn't a memorized analytical playbook — it's deep logical reasoning applied to each new problem.

From a business problem to a model

My core strength is translating a concrete business problem into something that can be computed. Asked to schedule ~500 people optimally across hotels, buses, and tours, I defined it as a quantity to minimize and wrote it as an optimization algorithm. The domain is interchangeable — moving people, staffing shifts, or making better use of existing resources to cut costs — because the real skill is the translation itself: turning a messy real-world goal into a model and an algorithm that solves it.

Catching costly errors others miss

When I analyze data or a process I keep asking: does this actually follow from the earlier definitions and data, or is it vague or forced? Is the process unambiguous? What are the consequences once it's deployed? Can it be simplified? Do we really need to collect this much information? This questioning catches expensive mistakes early.

• In a regulated environment, I caught a critical contradiction in documentation a team of specialists had written over several months — one part defined a process as recurring "every 3 months", another as "every 12 weeks", and three months is not twelve weeks.
• Reviewing payroll rules, I flagged an ambiguity in a collective labour agreement that referred to terms it had never clearly defined.

The same questioning shapes how I build ETL pipelines: I ask whether I really need to pull a given field or whether I can derive it from others — keeping the data flow lean and its logic sound.

Data integrity — your data stays safe

This is the same error-catching ability, applied under strict pharmaceutical rules. Remote bookkeeping demands trust in the safety and consistency of data — and that's where my background pays off. At Alvotech I worked under GMP and data-integrity rules: strict change control, documentation, and audit trails. I bring the same discipline to payroll:

• Zero transcription errors — the accountant's manual validations (proportional holiday pay, sick-leave limits, unusual overtime) become a repeatable, automated mechanism with no human copy step.
• Automated historical audit — algorithms scan the last 12 months to verify sick-leave limits (including children's sick days) before any payment, turning deviation analysis (root-cause thinking from regulated production) into payroll that resists anomalies.
• Local execution for sensitive data — the bespoke payroll calculation can run as local Python notebooks, so sensitive payroll data never leaves your own network.