Thomas René Decloedt

AI Engineer

Building AI solutions from idea to production.

About

Currently working at Jurimesh, where I own end-to-end delivery of AI solutions for complex document workflows.

MSc in Computer Science Engineering from Ghent University, graduated with distinction. My research was presented at NASA JPL’s onto Nexus Forum.

Experience

AI Engineer at Jurimesh

April 2025 – present

I lead Jurimesh's contract comparison feature end-to-end, enabling M&A, PE, and legal teams to compare large sets of contracts in minutes instead of days. The system highlights high-level deal risks as well as paragraph-level differences, supporting both strategic due diligence and detailed legal review.

My work spans the full product lifecycle: working closely with clients to understand workflows, researching and validating technical approaches, shipping production-ready MVPs, and owning the feature in production. The contract comparison system combines retrieval with targeted LLM analysis and task-specific processing to balance quality with scalability, and was featured in Jurimesh's launch video and blog post.

Beyond contract comparison, I designed Jurimesh's domain-specific chunking and document structuring pipeline for legal text, significantly improving retrieval quality across chat and comparison workflows while handling bursty workloads and highly variable document structures.

On the ML side, I fine-tuned multilingual clause classifiers and built robust retrieval-augmented generation (RAG) pipelines, with offline evaluation, automated regression testing, and LLM-as-a-judge pipelines to monitor and improve prompt and retrieval quality.

From an engineering perspective, I work across the stack: building distributed async worker systems for LLM workloads, production APIs, and real-time client features. This includes LLMOps, evaluation tooling, and production observability.

Python React TypeScript Vite Fastify PostgreSQL Drizzle Atlas Redis WebSockets GCP Vertex AI Qdrant Rust Kubernetes

Data Science Intern at TechWolf

Summer 2023

Time-series analysis, forecasting, and change point detection for emerging skills in vacancy data.

Python pandas scikit-learn

Software Engineering Intern at Sigasi

Summer 2022

Developed Java-based software solutions; gained exposure to software engineering best practices.

Java GitLab

Education

Master of Science in Engineering: Computer Science Engineering at Ghent University

2022 – 2025
  • Graduated with distinction
Python PyTorch Natural Language Processing Machine Learning Financial Mathematics

Bachelor of Science in Engineering: Computer Science Engineering at Ghent University

2018 – 2022
Python Java C C++

Projects

Distributed ML Training

Cloud-based distributed training setup using PyTorch DDP to train an MNIST CNN across multiple GCE VMs, with experiment tracking, model registry, and a web inference endpoint.

  • · 2-node distributed training (master + worker) with torchrun
  • · Metrics and hyperparams logged to Vertex AI Experiments
PyTorch DDP Vertex AI GCP Terraform FastAPI Cloud Run Docker uv

Foosball Rating Tracker

Full-stack foosball rating system on Cloudflare. Uses OpenSkill for multiplayer team-based rating (1v1, 1v2, 2v2).

  • · Real-time leaderboard, match history, player profiles, seasons
  • · Position-specific ratings (defense, attack, solo)
Cloudflare Workers Hono React Vite PostgreSQL Drizzle OpenSkill

Voice Agent

HR application with adversarial role-playing voice agent for conversation preparation.

  • · Technical lead in 3-person team; full-stack with Vue
  • · OpenAI Realtime Voice API integrated with LangGraph
Vue Vite TypeScript Python FastAPI PostgreSQL Redis Celery WebSockets OpenAI Realtime API LangGraph Docker GitLab CI

Mo-Lab (Master's Thesis)

Interactive model reporting and analysis in openCAESAR through Natural Language Processing and SPARQL integration. RAG-based text-to-SPARQL with domain adaptation for Model-Based Systems Engineering.

  • · Research presented to NASA JPL at onto Nexus Forum 2025
  • · RAG model with domain adaptation and synthetic data generation
Python PyTorch RAG GNNs LLM Fine-tuning QLoRA Synthetic Data SPARQL RDF Ontology Knowledge Graphs Domain Adaptation Jupyter AWS Flask