Remco Hendriks

AI & Machine Learning Engineer and Researcher

Developing intelligent systems through advanced machine learning, data engineering, and LLM-powered applications. Specializing in knowledge graphs, RAG architectures, and scalable ML infrastructure.

Based in Amsterdam, The Netherlands  •  Available for research collaboration & ML engineering projects

• View Experience
• Research & Publications
• Blog

My Expertise

With 15+ years in software engineering, I now focus on AI/ML systems that bridge research and production. My work spans the complete ML lifecycle—from data engineering and model development to deployment and monitoring.

Core Competencies

• LLM Systems & Applications: RAG/GraphRAG architectures, agentic systems, prompt optimization, LangChain (OpenAI, Anthropic, Llama)
• Machine Learning Engineering: Model training & fine-tuning (PEFT, LoRA), TensorFlow, PyTorch, MLOps pipelines, experiment tracking
• Data Engineering: ETL pipelines (Kafka, Airflow, Snowflake, Clickhouse), data transformation, feature engineering, knowledge graphs (Neo4J)
• ML Infrastructure: Model serving (Triton, vLLM), containerization (Docker/Kubernetes), CI/CD for ML, cloud platforms (AWS/Azure/GCP)
• Research & Development: SOTA research replication, novel architectures, experimental methods, performance optimization

Technical Stack

Python • PyTorch • TensorFlow • LangChain • TypeScript • Node.js • React/Vue • Neo4J • PostgreSQL • MongoDB • Docker • Kubernetes

Project:

Boehringer Ingelheim

Synthetic Persona System for Global Market Research

GraphRAG-based research tool for early-stage hypothesis testing through synthetic HCP interview replication, trained on authentic physician-interviewer transcripts across global markets.

• Custom fine-tuned models using PEFT (LoRA) trained on authentic HCP-interviewer transcripts from partner organizations, capturing real-world response patterns and communication styles
• RLHF (Reinforcement Learning from Human Feedback) ensuring synthetic personas replicate authentic interview dynamics observed in field research data
• Global knowledge integration spanning mainland China, Japan, EU, and USA healthcare markets, incorporating region-specific medical practices and market dynamics
• GraphRAG architecture processing extensive internal BI corpus data and medical literature into structured knowledge graphs for contextual response generation
• Hallucination mitigation through citation tracking, knowledge graph grounding, and validation against pharmaceutical documentation
• Generates synthetic Q&A transcripts modeling Healthcare Professional (HCP) and Payer organization perspectives for hypothesis development and scenario planning
• Enterprise deployment supporting internal market research teams with realistic interview simulations

Production LLM application demonstrating advanced fine-tuning techniques, GraphRAG architecture, and grounded generation methods to replicate authentic HCP interview transcripts, supporting pharmaceutical market research across global healthcare systems.

Project:

ExecutiveOrders.im

AI-Powered U.S. Executive Order Explorer

A specialized AI assistant for deep, topic-driven exploration of U.S. presidential executive orders, leveraging official Federal Register data. Public demonstration of GraphRAG techniques deployed in production for a law firm client.

• Developed an AI system utilizing cutting-edge GraphRAG (Microsoft Research) for superior information retrieval from complex legal documents
• Transforms unstructured legal texts into an intelligent, queryable knowledge graph—same architecture powering production legal research systems for case law, contracts, and regulatory document analysis
• Enables nuanced search beyond keyword matching, uncovering connections missed by traditional methods
• Achieves over 60% improvement in answer relevance and up to 90% reduction in LLM costs compared to standard RAG

ExecutiveOrders.im serves as a publicly accessible demonstration of enterprise GraphRAG techniques deployed for confidential legal clients, showcasing the application of advanced knowledge graph construction to complex legal document intelligence.

Visit: ExecutiveOrders.im

Project:

MyParcel

High-Performance Data Pipeline & Cloud Infrastructure

Processing 100,000+ daily shipments requires robust data engineering and real-time data processing capabilities.

• High-volume data pipeline architecture handling real-time parcel data ingestion, transformation, and indexing
• ElasticSearch implementation for sub-second search and retrieval across millions of shipment records
• LLM-powered ETL pipelines for automated data cleaning, validation, and correction of address and shipment information
• Event-driven microservices architecture for carrier API integrations (PostNL, DHL, UPS, FedEx, DPD)
• Scalable front-end infrastructure (Vue, Nuxt, Vite) optimized for high-frequency data operations
• Cloud deployment on AWS with auto-scaling capabilities

Re-architected the platform's data processing infrastructure, achieving 50% performance improvement and establishing ML-ready data pipelines for future predictive logistics features.

Visit: myparcel.nl

Project:

MatchWornShirt

Real-Time Data Processing & Recommendation System

High-frequency auction platform processing real-time bids with intelligent recommendations for sports memorabilia collectors.

• Real-time bidding engine with WebSocket architecture for sub-100ms bid propagation
• Redis-backed caching layer for high-throughput concurrent user sessions
• LLM-powered recommendation system for personalized item suggestions based on bidding history and preferences
• Structured data modeling and cataloging system for 1000+ player entities across 100+ sports teams
• Multi-currency and multi-language data transformation pipeline (12 languages, 10 currencies)
• Scalable Vue architecture deployed on AWS

Built real-time data streaming infrastructure and intelligent recommendation features, achieving 30% increase in user engagement through personalized ML-driven suggestions.

Visit: matchwornshirt.com

Project:

ABN AMRO

Enterprise Knowledge Graph & Automated Document Intelligence

Pre-LLM era knowledge extraction system processing 10,000+ enterprise documents to build a corporate knowledge graph.

• Automated knowledge graph construction from unstructured corporate documents, technical specifications, and architectural records
• Graph database architecture (Neo4J, CosmosDB) for modeling complex relationships across 5,000+ applications and 1,000+ business entities
• Neural word embedding models (Word2Vec) for semantic similarity and relationship extraction, production NLP pipeline with spaCy for named entity recognition and dependency parsing, establishing foundational techniques that inform modern transformer-based approaches
• Document parsing pipeline traversing 10,000+ technical documents to extract structured knowledge
• Graph traversal algorithms for data lineage visualization and impact analysis across enterprise systems
• Full-stack platform (Vue, Fastify) deployed on Azure Cloud

Pioneered pre-LLM knowledge graph construction techniques, demonstrating the power of graph-based knowledge representation for enterprise intelligence—foundational work that directly informs modern RAG architectures.

Visit: clarity.abnamro.com (internal)

Project:

Nationale Nederlanden

Multi-Tenant Enterprise Data Governance Platform

Enterprise-scale data management system for risk assessment and regulatory compliance across 5,000+ applications in 10 countries.

• Multi-tenant data architecture supporting complex role-based access control and data isolation requirements
• Automated compliance workflow engine for GDPR data processing and validation across international regulations
• Structured data collection and aggregation platform spanning 5,000+ enterprise applications
• Risk assessment data modeling and impact analysis pipeline for regulatory reporting
• Full-stack platform (Angular, Node.js, LoopbackJS) with multi-cloud deployment (AWS, GCP)

Built scalable data governance infrastructure for one of Europe's largest insurers, establishing enterprise-grade data management practices and automated compliance workflows.

Visit: seca.insim.biz (internal)

Project:

vatfree.com

VAT Refund Platform with Automated Document Processing

Mobile refund platform processing 1,000+ daily claims across 5,000+ EU retail stores.

• Cross-platform mobile application (React Native) for refund claim submission and processing
• Computer vision and OCR implementation for receipt scanning and automated data extraction—demonstrating pre-LLM AI/ML experience in production document processing
• Real-time VAT calculation engine with multi-currency support and validation
• Store integration system covering 5,000+ retailers across the EU
• Backoffice data management platform (Angular, Express) handling claim processing workflow
• Multi-cloud deployment (AWS, GCP)

Built end-to-end refund processing platform, including applied computer vision for document automation, establishing practical AI/ML experience before the transformer era.

Visit: vatfree.com

Project:

ING

Enterprise Payment Gateway Processing 800K+ Monthly Transactions

High-volume payment processing infrastructure for SME merchants with secure VISA and Mastercard integration.

• End-to-end payment gateway development handling real-time VISA and Mastercard authorization, clearing, and settlement workflows
• High-throughput transaction processing pipeline managing 800,000+ monthly transactions across 1,000+ merchant integrations
• Supervised ML fraud detection using logistic regression with geospatial feature engineering—cardholder-merchant location correlation, cross-border transaction pattern analysis, and historical behavior modeling per card and merchant for real-time risk scoring
• PCI-DSS compliant data handling architecture with 3D Secure implementation for payment authentication
• Low-latency API design for seamless merchant integration with webshops and applications
• Multi-language development stack (Angular, Node.js, Python) deployed on AWS

Built mission-critical payment infrastructure with production machine learning for fraud detection, demonstrating classical ML techniques in high-stakes financial systems before the deep learning era.

Research & Publications

My research interests focus on democratizing machine learning systems, distributed computing architectures, and advancing practical ML infrastructure for heterogeneous environments.

Selected Publications

"MLitB: Machine Learning in the Browser"
E. Meeds, R. Hendriks, S. Al Faraby, M. Bruntink, M. Welling
PeerJ Computer Science, 2015 | DOI: 10.7717/peerj-cs.11

• Pioneered browser-based distributed ML framework enabling training across heterogeneous devices without installation—anticipating modern federated learning paradigms by nearly a decade
• Developed adaptive workload distribution system that dynamically allocates computation based on device capabilities (processing power, network bandwidth, memory constraints)—foundational work now critical in SOTA distributed LLM training
• Designed stochastic synchronization protocol handling devices with 100x performance variation, from mobile phones to workstations, within unified training framework
• Demonstrated distributed SGD on heterogeneous networks scaling to 64+ nodes, establishing early techniques for training on unreliable, variable-capability devices

Read Paper (PeerJ) | arXiv | Google Scholar

About me

My work spans machine learning research and engineering, with over 15 years building production systems. I architect AI/ML solutions—from distributed training infrastructure and knowledge graphs to LLM-powered applications—with a foundation in full-stack development that ensures practical, deployable systems.

Beyond client work, I actively pursue R&D in distributed ML, agentic systems, and advanced RAG architectures, exploring techniques that push the boundaries of what's possible with modern AI.

Based in Amsterdam with frequent time in Asia (China, Taiwan, Japan). When not building AI systems, I'm advancing my Mandarin Chinese (HSK6), discovering great food, or exploring by bike.

Research & Engineering Services

Available for collaboration on AI/ML research and production systems:

• AI Chatbots & Conversational AI: Intelligent chatbots, virtual assistants, AI personas using LLMs (OpenAI GPT, Anthropic Claude, Llama)
• LLM Applications & AI Agents: RAG/GraphRAG architectures, agentic systems, production LLM deployments
• Distributed ML Systems: Training infrastructure, heterogeneous computing, model parallelization
• ML Engineering: Model development, fine-tuning (PEFT/LoRA), MLOps pipelines, experiment tracking
• Data Engineering for AI: ETL pipelines (Kafka, Airflow, Snowflake), knowledge graphs, feature engineering
• Full-Stack Web Applications: Production web and mobile applications (React, Vue, Node.js, Python)—proven enterprise track record
• Research Collaboration: SOTA replication, distributed learning algorithms, novel architectures
• AI Strategy & Advisory: Technical consulting, system architecture, ML infrastructure design

Based in Amsterdam, The Netherlands. Available for consultancy, research collaboration, and project development.