Hey there,

I'm Jelte Oldenhof

A Data Scientist and Full-Stack Data Analyst with an edge in Electrical Engineering.

Bridging Engineering, AI and Product-Oriented Data Systems.

MSc Data Science, Leiden University, graduating August 2026
Available full-time from October 2026

Portrait of Jelte Oldenhof

Education

MSc in Computer Science: Data Science (incl. Pre-Master)

2024 - Aug 2026 (expected)

Leiden University, The Netherlands

MSc thesis: graph-based anomaly detection in ERP accounting data using an unsupervised Graph Autoencoder with a GraphSAGE encoder, plus explainability methods for transparent, auditable anomaly interpretation.

Selected Coursework

  • Deep Learning
  • Cloud Computing
  • Reinforcement Learning
  • AutoML
  • Social Network Analysis
  • System & Software Security
  • Software Development & Product Management

2024 - 2026

BSc in Electrical Engineering, Minor in Embedded Systems

2016 - 2020

Hogeschool Rotterdam, The Netherlands

2016 - 2020

Professional Experience

MACH Technology, The Netherlands

2023 - Present

Full-Stack Data Analyst (part-time, alongside MSc)

Designing, building and maintaining internal data and web applications (React/Flask), including a tool used by ~8 stakeholders to visualize demand and idle stock across two production sites. Applying machine learning to classify component demand patterns and modelling component relationships as a network (Python, NetworkX) on top of SQL Server data. Collaborating across procurement, sales support, work planners, IT and R&D.

2023 - Present

MACH Technology, The Netherlands

2020 - 2023

Hardware & Embedded Software Engineer R&D

Delivered around ten prototypes and final products, such as home trainers, check-in systems and connected devices, covering the full cycle from firmware to PCB design and manufacturing. Worked directly with clients on requirements engineering, acting as product owner on several projects.

2020 - 2023

MSc Thesis

“The Irregular Neighbourhood”: Explainable Graph-Based Anomaly Detection in ERP Accounting Data

Graph Neural Networks

Leiden University, inspired by an industry research proposal from KPMG NL

Graph Neural Networks

For forensic audit, detecting an anomaly is not enough: every flagged transaction needs transparent, defensible evidence. In my thesis I designed a reproducible pipeline that transforms raw journal entries into a heterogeneous graph connecting entries, general-ledger accounts and profit centers, and scores each entry with an unsupervised Graph Autoencoder (GraphSAGE encoder). A built-in explainability layer answers the two questions an auditor actually asks: which relationship made this entry suspicious, and which field. The pipeline outperforms both tabular and graph baselines and detects relational anomalies that are invisible to tabular models.

Highlights

  • Heterogeneous graph of 533K nodes and 2.1M edges, built from raw ERP journal entries
  • Unsupervised GraphSAGE autoencoder with a leak-free three-phase evaluation protocol
  • Detects local anomalies that tabular models cannot see, using relational context alone
  • Integrated explainability: relational (occlusion) and feature-level evidence per flagged entry
  • Validated on a synthetic SAP FI/CO benchmark and real-world vendor payment data

Case Study

Data-Driven Inventory Optimization

Full-Stack Data Analysis

Strategic Sourcing Team

Data Analysis

In my current project, I built a full-stack web application that interfaces with our company's MRP systems to drive smarter inventory decisions. By extracting data with SQL, mapping complex relationships between components using Python’s NetworkX library, and recommending viable alternatives, my work helps reduce slow-moving and obsolete parts. This initiative not only optimizes inventory but also supports a lean, just-in-time strategy across production sites.

Highlights

  • SQL-driven data extraction from MRP systems
  • Network modeling with Python’s NetworkX
  • Development of a full-stack web application
  • Strategic sourcing to reduce inventory of obsolete parts

Skills & Toolkit

Programming

Python, SQL, C/C++, TypeScript, React, Flask, Git, MATLAB, LaTeX, Linux/SSH

Data Science

Machine learning, deep learning, graph neural networks (GraphSAGE, GAT, autoencoders), anomaly detection, explainable AI, statistical modelling, forecasting

Cloud & Methods

Microsoft Azure (Functions, App Service, Blob Storage), containers, LLM/API integration, Agile/Scrum, JIRA, MRP systems

Engineering

Embedded systems design, PCB design, wireless communication, low-level protocols, requirements engineering, firmware

Languages

Dutch (native), English (C1 advanced), German (beginner, actively learning)

Interests

Outside of work, you’ll often find me challenging myself on the climbing wall or out on a long mountain hike. Whether it’s an indoor climb or a rugged mountain trail, I see each ascent as a puzzle that challenges both my body and my mind.

I also love camping, spending nights under the stars and enjoying the simplicity. These activities keep me grounded, boost my creativity, and remind me why I love tackling challenges every day.

Background

Data and Tech

Technology has been a part of my life since I was little, thanks to my dad and his world of electronics. My fascination with physics naturally led me to study Electrical Engineering, and my early career in embedded systems taught me the ropes.

Somewhere along the way, I discovered the true power of data, which really changed the game for me. Combined with my growing interest in business values and a hunger for bigger challenges, I kicked off my Master’s in Data Science.

And here I am today, living in the fast lane of data and tech, with a lot of passion and motivation. I embrace all the challenges and work on projects I never thought possible, like training a graph neural network to find anomalies in half a million accounting records for my thesis, while running my own cloud data pipeline on the side. I'm enjoying every twist and turn alongside an awesome community of friends who share the same passion.

Other Projects

Switchback: Cloud-Deployed Automated Data Pipeline

Cloud & LLM Engineering

End-to-end pipeline on Microsoft Azure that ingests data from multiple public APIs, deduplicates by content hash, and scores relevance with an LLM (Anthropic Claude) using prompt caching and strict-JSON output to bound cost. Runs live with a Flask dashboard, scheduled daily runs, email alerts and hardened authentication.

Cloud & LLM Engineering

Github Repository: Data Science in Social Networks

Data Science in Social Networks

A comparison study on Node Embeddings for the Social Network Science course at Leiden University.

Data Science

Requirements Engineering Process

Requirements Engineering

Enhanced requirements engineering standards at MACH Technology using the Volere Method.

Requirements Engineering

Smart Wireless Communication Switch

Embedded Systems

Developed prototypes for a 400VAC switch controlled wirelessly via the OOK (On/Off Keying) protocol.

Embedded Systems

High Power Motor Controller

Embedded Systems

Developed C code for a high-power BLDC motor and created a matching Python-based GUI.

Embedded Systems

Hardware Design For Manufacturing

Hardware Engineering

Designed printed circuit boards (PCB) for manufacturing using Altium software (2D and 3D tools).

Hardware Engineering

Contact Me

Jelte Oldenhof on LinkedIn

Feel free to reach out for collaborations. Open for full-time data science roles from October 2026.