Roman Slaap

Agentic AI Engineer | Forward-deployed delivery for enterprise clients

📍 Haarlem, Netherlands

✉️ romanslaap@gmail.com

📱 0653551566

Resume

About Me

For me, engineering is understanding the underlying problems first. I talk to the people involved and affected, understand their issues and wishes, and only then start designing and implementing a solution. With over 8 years of experience working with large enterprises in retail, automotive and insurance, I understand how they decide, build and ship, and where and how it tends to fail. I deliver well-tested systems that other engineers can understand and work with, and that run efficiently in production through monitoring, evaluation and cost optimization.

Testimonials

"[Roman] never focused solely on writing code or solving isolated technical challenges. Instead, he consistently considered the complete solution, from understanding the business problem and technical architecture to implementation, deployment, and long-term maintainability. [...] He is someone you can trust to take responsibility for a problem, work independently, and deliver high-quality solutions without requiring close supervision."

Global Head of AI & Tech, Inter IKEA Group

Radmilo Krstonijevic

Experience

AI Engineer

Machine Learning Engineer & Consultant

Machine Learning Engineer

Data Scientist/Engineer

Time-series Analysis Intern

AI Engineer

March 2024 - September 2026

Inter IKEA Group | Roman Slaap Data & AI Consultancy (self-employed)

Built and managed an agentic AI platform, piloted with 40 regular users in 2 stores and reaching over 100 unique users in total, cutting the time to find the information they needed from 1 to 3 minutes of manual searching to about 30 seconds.
Hired as the ML and generative AI expert and technical decision maker, choosing the technology stack and models.
Built evaluation sets from user feedback, frequent questions and the questions experts deemed important to get right, and used them with observability and prompt optimization to measure response quality and improve the agent over time.
Built a self-service platform where managers create and manage their own agents, with a dashboard that tracks the questions coworkers ask and how they rate the answers.
Designed a modular agentic architecture connecting 8+ knowledge sources behind one interface, working with stakeholders to prioritize which sources to integrate first.
Owned the platform and its server infrastructure end to end, keeping total running costs to about €200 a month (about €5 per regular user) while maintaining performance and reliability.

Skills Developed and Used

Python FastAPI Terraform LangGraph MCP Azure AI Search Observability (Langfuse, MLflow) Databricks Authentication & authorization GitHub Actions Agentic architecture RAG LangChain Vector databases Prompt engineering Graph-based data ingestion LLM evaluation Cost optimization API development JavaScript System architecture ETL pipelines PySpark Git Azure Stakeholder management MLflow Testing (pytest) Docker TypeScript SvelteKit

Machine Learning Engineer & Consultant

May 2022 - February 2024

Thoughtworks, Amsterdam

Wrote 2 technical proposals for prospective clients; both became signed statements of work, leading to the Mondelez and Mercedes-Benz projects.
Team lead for 2 junior engineers on the Mercedes-Benz project, coaching them in software and ML engineering; one became a cloud engineer at Google, the other a senior developer at Thoughtworks.
Led the office's biweekly lunch and learn program with internal and external speakers, and organized and hosted the MLOps & Crafts meetup (2 workshops, about 50 attendees each).
Gave the Data Mesh keynote with Thoughtworks' head of engineering at Blipz on the Radar 2022 (tech conference, Utrecht), and ran trainings for colleagues on software engineering best practices and ML engineering fundamentals.

Skills Developed and Used

Stakeholder management Client acquisition Proposal writing Technical presenting Coaching & mentoring Community building

Machine Learning Engineer

April 2023 - November 2023

Mercedes-Benz (via Thoughtworks), Stuttgart (remote)

Took a proof of concept to a production system on AWS in 6 months, giving pricing analysts near real-time competitor prices in place of manual lookups on competitor websites and in PDFs.
Designed the model architecture and built the pipeline end to end, from gathering unstructured competitor data (websites and PDFs) through processing and modeling to automated retraining and monitoring.
Raised car-part matching accuracy from 30% to over 90% on a hold-out test set within 2 months, using multiple rounds of analyst corrections fed into automated retraining.

Skills Developed and Used

Python AWS AWS SageMaker Docker GitLab CI Predictive modeling Human-in-the-loop ML Automated retraining Model monitoring System architecture

Machine Learning Engineer

October 2022 - February 2023

Mondelez (via Thoughtworks)

Built generative and Bayesian optimization models, with Gaussian processes choosing which recipes to test next, so product experts needed fewer test rounds to optimize recipes from under 20 data points.
Combined a generative adversarial architecture with human-in-the-loop input from experts and consumer test panels.

Skills Developed and Used

Python Probabilistic modeling (Gaussian processes) Bayesian optimization (BoTorch) Generative & adversarial models Modeling with small datasets Human-in-the-loop ML Azure CI/CD

Data Scientist/Engineer

January 2021 - 2022

Nationale Nederlanden, Rotterdam

Built models that forecast the probability and size of insurance damages (claim frequency and severity), backtested on later time periods than they were trained on (out-of-time validation), used to set risk premiums and commercial prices across multiple product lines under strict regulatory constraints.
Built CI/CD deployment pipelines to support continuous model development and release.
Automated manual processes, including contract data extraction, reducing manual effort.
Built dashboards and visualizations to turn raw data into decision-ready information for business stakeholders.

Skills Developed and Used

Python Forecasting Backtesting & out-of-time validation GLMs Gradient boosting (XGBoost) scikit-learn SQL Azure CI/CD Statistical modeling ARM templates

Data Scientist

2020 - 2021

Nationale Nederlanden, The Hague

Led the Rating Engine project, designing and building an Azure-based end-to-end pricing and model management system.
Applied statistical and predictive modeling under strong regulatory and business constraints.
Supported pricing teams with data-driven recommendations on damage insurance pricing policy.
Led the data science chapter, coordinating knowledge sharing between 3 teams and 15 data scientists through regular sessions and shared coding and modeling standards.

Skills Developed and Used

Python SQL Azure Statistical modeling Predictive modeling Coaching & mentoring Community building

Pricing Analyst

2019 - 2020

Nationale Nederlanden, The Hague

Built risk-premium, commercial pricing, and churn models across multiple insurance product lines using statistical and predictive modeling techniques.

Skills Developed and Used

Python Statistical modeling Predictive modeling

Time-series Analysis Intern

September 2017 - August 2018

ING, Amsterdam

Anomaly detection in transaction-based data
Thesis internship

Skills Developed and Used

Python Forecasting Time-series analysis LSTM Anomaly detection PII data handling

Education

MSc Computational Science

Universiteit van Amsterdam

2015 - 2018

Minor: High Performance Computing

National University of Singapore

September 2016 - January 2017

BSc Artificial Intelligence

Universiteit van Amsterdam

2011 - 2015

Certifications

Data Engineering Accelerated

Thoughtworks

2023

Tech Lead Development Programme

Thoughtworks

2023

Data Management for Data-Stewards

Vijfhart IT-opleidingen

December 2021

Azure Developer Associate

Microsoft

AI-102, September 2021

Azure AI Engineer Associate

Microsoft

AI-102, November 2020

Azure Fundamentals

Microsoft

Az-900, October 2020

Skills

LLM & Agent Systems

Agentic architecture RAG MCP LangGraph LangChain Azure AI Search Vector databases Prompt engineering

Evaluation & Monitoring

LLM evaluation Observability (Langfuse, MLflow) Model monitoring Cost optimization

ML Engineering & MLOps

Human-in-the-loop ML Automated retraining MLflow AWS SageMaker

Data Science & Statistics

Forecasting Time-series analysis Backtesting & out-of-time validation Statistical modeling Predictive modeling GLMs Gradient boosting (XGBoost) Probabilistic modeling (Gaussian processes) scikit-learn Bayesian optimization (BoTorch) Generative & adversarial models Modeling with small datasets Anomaly detection LSTM

Data Engineering

SQL PySpark ETL pipelines Databricks PII data handling Graph-based data ingestion

Software Engineering

Python Testing (pytest) FastAPI API development Authentication & authorization System architecture CI/CD GitHub Actions GitLab CI Docker Git TypeScript JavaScript SvelteKit

Cloud

Azure AWS Terraform ARM templates

Consulting & Leadership

Stakeholder management Client acquisition Proposal writing Technical presenting Coaching & mentoring Community building

Languages

Dutch (Fluent) English (Fluent) Portuguese (Beginner)