Hi, I'm Rasmus Kampmann
Data Analyst | BI Specialist | Power BI Developer
- Power BI
- SQL
- Microsoft Fabric
- Python
I turn scattered commercial data into systems you can trust.
3+ years working in commercial, now building the reporting a business runs on, and the data foundation AI gets layered on top of. Semantic models, dashboards, forecasting, in agriculture and beyond.
What I Do
Power BI and SQL are most of the work. The rest is what the work needs.
Power BI
I build the semantic model first, then the dashboard on top. DAX, Power Query, and a model that stays maintainable, so reporting drives decisions instead of just showing numbers.
SQL
I build the data layer that everything else runs on. Data modeling, ETL, PostgreSQL. Raw source data in, analysis-ready tables out, and the team can query them without asking me first.
Microsoft Fabric
Pipelines, Python notebooks for forecasting, and Fabric apps. Ingestion, modelling, and reporting in one platform instead of three tools that don't talk to each other.
Python
I build the automation and the models spreadsheets can't handle. The ML work runs on scikit-learn and pandas. The pipelines run on a schedule instead of by hand.
Excel
Financial modelling belongs in Excel. Forecasts, budgets, margin and customer profitability. KPI definitions that hold across sales, operations, and production.
AI in the data flow
AI goes on top of a data foundation, not instead of one. I use Claude Code and Copilot inside the work itself: data prep, SQL and DAX, documentation. A tool in the stack, not the headline.
My Projects
Production systems behind reporting, forecasting, and commercial decisions. Written for non-technical readers.
The reporting a business runs on
Real client systems at an agricultural seed business. One source of truth reconciled to the accounts, and a production plan that stays live instead of going stale on every sale.
Invoice & Financial Dashboard
One trusted view of profit per product, profit per customer, and the cash owed, built from invoices and reconciled to the accounts within 1.25%. Commercial figures below are illustrative; the reconciliation is real.
Operations & Production Planning
A live system that tells a seed business what to produce, how much, and when. It reproduces the planner's own numbers exactly, validated against their spreadsheet. A planning system with scenario testing, not a forecast.
Analysis on a model you can trust
The data model comes before the finding. Star schema in SQL, validated in DuckDB, and only then a number worth acting on. Synthetic data, real method.
Funnel & Segment Analysis
Funnel decomposition and firmographic segment-lift analysis across 102,007 records. One employee band converted at 37.5%, everyone else at 3.0%. The drop was after the first conversation, not before it. Synthetic data, real method.
Channel & Churn Analysis
Multi-factor scoring and twelve-month survival analysis across ten channels reported as one blended mix. Scored on win rate, capacity, speed, and churn. Synthetic data, real method.
What AI can do once the foundation is there
Both of these are AI sitting on a data layer I had to build first. The registry features and the ingestion corpus are what these models stand on.
Software Detection ML Model
A model is only as good as the features under it. This one predicts accounting-software adoption from public Danish company-registry data alone, no paid data vendor, and reorders the cold-outbound dialler queue before reps make a single call.
Agentic External Intelligence Platform
The RAG chat is the part you see. Underneath it sits a 228-source ingestion layer with a 7-layer scraper fallback. No corpus, no citations, no signal. Personalised research, competitor monitoring, and weekly briefings for a B2B agriculture company.
About Me
Hi, I'm Rasmus. I turn scattered commercial data into systems you can trust.
I build the reporting a business runs on, and the data foundation AI gets layered on top of. One source of truth, semantic models, dashboards, forecasting.
The problems I get called in for:
- ERP, CRM, and spreadsheet data that don't agree
- KPIs defined differently in every report
- Reporting rebuilt by hand every week
- Forecasting done on gut feel
- AI on the roadmap with no data foundation under it
My background is in commercial operations, across sales, marketing, finance, and the day-to-day of running a business. That's what sets the work apart: I read the numbers from inside the business, so I know what they mean, where reporting breaks, and how sales, finance, and operations actually use them, because I've worked in those functions, not just reported on them.
At Veginova, an agricultural seed business, I'm building the data, BI, and reporting the business runs on: the finance source of truth, the production planning engine, and the forecasting leadership decides on.
How I work:
- Build clean, consistent data structures as the foundation
- Automate ingestion and transformation instead of rebuilding manually
- Keep pipelines lightweight, not over-engineered
- Build Power BI models that are easy to maintain and extend
- Build dashboards around the decisions leadership makes
- Use AI inside the data flow, where it makes the work faster or more accurate
What I've delivered:
- Standardised KPI definitions across sales, operations, and production. Accuracy up 40%+.
- Automated reporting pipelines. 10+ hours of manual work cut per week.
- Forecasting built into a production planning system that sets what to produce, how much, and when.
- Financial data unified into a profitability dashboard, reconciled to the accounts.
Stack:
Power BI: semantic models, DAX, Power Query.
SQL: data modeling, ETL, PostgreSQL.
Microsoft Fabric: pipelines, Python notebooks for forecasting, Fabric apps.
Excel: financial modelling, forecasts, budgets, margin.
AI in the data flow: Claude Code, Copilot, and LLM workflows for data prep, SQL and DAX, and documentation.
Hands-on across ERPs, CRMs, and spreadsheet/BI tools: the full commercial data layer.
Most of my experience is from small companies and my own. Close to the decisions, owning the work end to end.
I replace manual spreadsheets and disconnected reporting with systems that run on their own and make the business easier to understand. The outcome teams hire me for: stop rebuilding reports every Monday, stop questioning the numbers, start making commercial decisions from data you trust.
Danish, English, Spanish.
Open to data, BI, and Power BI developer roles in Denmark.
Experience
Roles where I built the data systems, dashboards, and automation behind real commercial decisions.
Jun 2026 – Present · Full-time
Data Analyst & BI Specialist
Veginova Seeds
- Back at Veginova, building the data, BI, and reporting the business runs on. Extends the reporting stack I already owned here into the financial and operational layer.
- Built the finance source of truth from invoice lines, reconciled to the official accounts, so profit per product and profit per customer are visible for the first time.
- Built the production planning engine that sets what to produce, how much, and when, on a one-year seed lead time. It reproduces the planner's own numbers exactly and lets them test a scenario before committing seed.
- Brought the scoring, forecasting, and revenue analysis I built at Digi-Tal into the finance and operations model here.
Stack: Power BI · SQL · PostgreSQL · Python · Supabase · Claude Code
Feb 2026 – May 2026 · Full-time · Hybrid
Data Analyst & Commercial Analytics
Digi-Tal Regnskab
- Owned the data work behind a Danish SMB accounting and fintech firm: ICP, scoring, software detection, and full-channel revenue analysis, from raw source through to operationalised scores feeding sales.
- Built ICP and predictive lead-scoring models plus an ML classifier identifying a prospect's accounting software (holdout AUC 0.75, permutation test p < 0.0001) from Playwright scraping and enrichment waterfalls.
- Full-channel revenue analysis across Meta, Google, LinkedIn outbound, cold calling, and inbound: close rate, conversion, meeting time, and pipeline velocity by channel and segment. Identified which channels drove customers and which burned budget.
- Customer, churn, and attribution analysis: segmented the network into customer, lead, lost, and inactive, and built a full-funnel attribution model joining campaign data, CRM deals, and the Danish business registry to isolate the signals separating payers from non-converters.
- Rebuilt the sales commission model with forecasting weightings, shifting payouts toward subscription and MRR so sellers optimised for LTV instead of discounting to close. Operationalised the scores through an end-to-end LinkedIn pipeline (scraping, enrichment, scoring, sequencing).
Stack: SQL · Python · Pipedrive · Clay · Playwright · Apify · HeyReach · Claude
May 2025 – Feb 2026 · Full-time · Hybrid
Data Analyst & BI Specialist
Veginova Seeds
- Owned the BI and reporting stack across sales, operations, and production. Consolidated the source data and stabilised the reporting the team ran on.
- Improved KPI accuracy by 40%+. Standardised metric definitions across teams.
- Cut reporting time by 10+ hours per week. Stabilised broken reporting workflows and consolidated data sources.
- Resolved data inconsistencies across inventory, sales, and production systems.
Stack: SQL · Power BI · Python · Excel · Clay · Claude Code
Aug 2023 – May 2025 · Full-time · Hybrid
Marketing Specialist & RevOps
Veginova Seeds
- B2B marketing and RevOps in agriculture, working with international wholesale, distributor, and grower customers.
- Outbound lead sourcing, CRM hygiene, and lead scoring and prioritisation.
- Trade-fair and grower-event sourcing, pipeline reporting, and channel attribution.
Jun 2024 – Aug 2025 · Self-employed · Remote
Founder · Data & RevOps
Sira Logic
- Service business building AI-driven lead generation, enrichment, and CRM automation for B2B companies.
- Built lead enrichment and scoring workflows. Qualification accuracy improved by 30–40%.
- Built CRM automation pipelines integrating HubSpot and GoHighLevel with external data sources.
- Built custom web scraping systems for industry-specific data sources.
Stack: SQL · Python · Clay · Apify · HubSpot · GoHighLevel · HeyReach
Jan 2025 – Jan 2026 · Freelance · Remote
AI & LLM Data Analyst
Outlier
- Reviewed and annotated 1,000+ Danish-language AI conversations to improve response quality.
- Spotted patterns where models failed and fed that back into training data.
- Built practical understanding of how LLMs work and where they break. The same insight powers the AI scoring and qualification workflows I build today.
Jan 2023 – Aug 2023 · Part-time · Hybrid
Marketing & Digital Graduate
Damstahl Danmark
- Managed product and marketing data in ERP and CRM platforms across European markets.
- Built Excel dashboards for campaign reporting.
- Standardised regional data processes.
See my freelance services
View Services →Recruiters: see my LinkedIn or grab my CV in English or Danish.