Klinkende Taal

Overview

I led the development of AI systems to help Dutch government organisations communicate more clearly with citizens. The core challenge was fine-tuning large language models to produce text at B1 reading level - making letters from the government accessible to 80% of the population.

Klinkende Taal / SlimTaal, 2024-2025

Technical Achievements

Model Evaluation & Selection

I designed and implemented a comprehensive evaluation system to identify the optimal base model for fine-tuning:

• Built a multi-service evaluation platform (Java/Spring Boot + Kotlin) comparing 36 LLMs across 4 providers (Mistral, OpenAI, Groq, Nebius)
• Developed an LLM-as-judge framework with 4 specialized judges evaluating data correctness, style guide compliance, quality adherence, and content rules
• Integrated the Klinkende Taal API for automated B1/B2 language complexity scoring
• Achieved 75.6% B1 compliance with Mistral-Small-24B + DPO, up from 12.4% baseline - a 63 percentage point improvement

Fine-Tuning Pipeline Development

I developed a complete ML pipeline for training language models on simplified Dutch:

• Implemented three-phase training: Domain Adaptation, Supervised Fine-Tuning (SFT), and Direct Preference Optimization (DPO)
• Built async data pipelines processing 160k+ Wikipedia sentences through B1 filtering and LLM quality assessment
• Developed curriculum learning approaches with 43k preference pairs sorted by difficulty gap
• Deployed training infrastructure on Nebius H100 80GB GPUs using Axolotl framework
• Created SGLang and vLLM inference pipelines for high-throughput model evaluation

Synthetic Data Generation

Real training data for government correspondence is privacy-sensitive, requiring synthetic alternatives:

• Built LetterProcessing system for standardizing and pseudonymizing correspondence templates
• Developed SyntheticLetters pipeline generating DPO training pairs from templates
• Created 36k combined training examples (32k sentence pairs + 5.3k letter pairs)
• Implemented quality filtering with automatic rejection of ambiguous or low-quality pairs

Human Feedback Platform

To continuously improve model outputs, I built the LveRLHF expert feedback system:

• Developed a gamified web platform for language experts to evaluate model outputs
• Implemented Kubernetes deployment with secure authentication for domain-restricted access
• Created API infrastructure for batch loading evaluation data and collecting feedback
• Deployed at expertportaal.slimtaal.nl for production use by Lve language specialists

Key Contributions

• Established the complete ML infrastructure for B1 language model development
• Identified Mistral-Small-24B as optimal base model through rigorous comparative study
• Built reusable evaluation frameworks with objective metrics for language complexity
• Created comprehensive documentation enabling knowledge transfer to the wider team
• Developed privacy-preserving synthetic data generation for sensitive government correspondence

Technologies & Skills

• Python, Java, Kotlin, Spring Boot
• PyTorch, Axolotl, vLLM, SGLang
• Direct Preference Optimization (DPO) and Supervised Fine-Tuning (SFT)
• H100 GPU training on Nebius cloud
• PostgreSQL, Flyway migrations
• LLM APIs (Mistral, OpenAI, Groq, Nebius)
• Kubernetes, Docker
• Async data processing pipelines