Data & AI
Five Legged
Data Sheep
Experts in all facets of data projects, from source to dashboard, including modeling and analytics.
bellen.ai
live ↗Development of a production-ready AI voice agent platform for restaurants — where generative AI, LLMs and modern software engineering come together.
What I build
- ▹ Designing and building LLM-powered AI assistants for customer interaction.
- ▹ Building agentic AI workflows with tool calling, memory and RAG for reservations, FAQs and business processes.
- ▹ Integrating AI services with telephony, databases, calendars and external APIs.
- ▹ Developing scalable backend services and AI APIs in Node.js with cloud-native deployment.
- ▹ Setting up AI engineering workflows, CI/CD, monitoring and evaluation of AI systems.
- ▹ Designing reusable AI architectures focused on reliability, maintainability and production use.
- ▹ Intensive use of OpenAI, Anthropic, Gemini and local LLMs across production and development.
AI voice agent · realtime
Tech stack
Result
Live at bellen.ai — a production AI voice platform handling real restaurant calls.
02
Residual Mix World Map
live ↗Interactive world map visualising the certified and residual energy mix for countries worldwide, including carbon intensity. Built for residualmix.org — users can explore tracked renewables, fossil generation, and CO₂ emission factors by country and year.
Tech stack
Data story
Input
AIB energy certificate data (CSV) covering European and global markets by country, year, and technology
↓
Output
Choropleth world map with three switchable KPIs: tracked generation, tracked renewables, and residual mix carbon intensity
↓
Result
Live at residualmix.org — interactive map serving energy transparency data across 50+ countries
03
Open Source Data Stack
live ↗Building a data warehouse backend for a web app that serves audio and video from the rich archive of the Netherlands Institute for Sound and Vision.
Tech stack
Data story
Input
Audio/video archive metadata from multiple source systems at the Netherlands Institute for Sound and Vision
↓
Output
Normalised, indexed documents in Elasticsearch powering archive search and discovery
↓
Result
Live webapp served by data backend at https://www.beeldengeluid.nl/schatkamer
04
Azure Data Warehousing
Designed and built a data warehouse at Syntrus Achmea Hypotheken, ingesting 10+ operational data streams into a unified model serving mortgage finance reporting.
Tech stack
Data story
Input
10+ operational data streams from mortgage finance systems (loans, valuations, payments)
↓
Output
Unified DWH model serving 50+ mortgage finance reports
↓
Result
50+ reports unified under a single governed data model
05
AI & Machine Learning Models
Led a data science team at Van Lanschot Kempen delivering churn and default prediction, M&A prediction, NLP-powered lead generation, and graph network visualisations.
Tech stack
Data story
Input
Client transaction history, financial data, unstructured text, internal + external data sources
↓
Output
Churn scores, default risk predictions, NLP lead rankings, graph network visualisations
↓
Result
Improved customer satisfaction. Better lead conversion.
About
Steven is a Data & Analytics expert with a deep understanding of Data Engineering, based in the Netherlands. He designs and builds data warehouses, machine learning systems, and open-source data stacks for financial services and media organisations. Currently building the data platform for the Netherlands Institute for Sound and Vision.
Skills