Hello, I'm

Majid Lotfian Delouee

I'm a  |

Ph.D. Computer Science  ·  Postdoctoral Researcher, Amsterdam UMC  ·  70+ citations  ·  EU citizen

Scroll

About Me

I'm a researcher and ML engineer with a Ph.D. in Computer Science from the University of Groningen (2024), currently working as a Postdoctoral Researcher at Amsterdam UMC's Translational AI Lab. My work covers pre-training foundation models from scratch, building clinical ML pipelines, and applying federated learning to biomedical problems.

During my PhD I worked on real-time complex event processing, privacy-preserving stream systems, and LLM-assisted rule generation for high-volume data streams using Apache Kafka and Flink. At Amsterdam UMC, I train and evaluate transformer-based models on clinical datasets with 1.2M+ records, in collaboration with the University of Cambridge.

My work sits at the intersection of research and engineering: I design model architectures and evaluation frameworks, and I build the pipelines and MLOps workflows that put them into practice. I hold research affiliations with the University of Amsterdam, University of Oslo, and TU Ilmenau.

 Almere, Netherlands  EU Citizen · No visa required

70+ Citations
14+ Publications
6+ Years Research
5 Institutions

Research & Expertise

Foundation Models & LLMs

Pre-training transformer models from scratch, fine-tuning, PEFT/LoRA, domain adaptation, and LLM-assisted system design for structured tasks.

Clinical AI & Healthcare

ML pipelines on large-scale clinical datasets, patient classification, biomarker discovery, and clinical decision support systems.

Stream Processing & CEP

Real-time event processing with Apache Kafka and Flink, complex event pattern detection, and high-throughput distributed pipelines.

Federated Learning

Privacy-preserving distributed training across institutions, federated evaluation frameworks, embedding-based FL, and governance mechanisms.

Anomaly Detection

Pattern and anomaly detection in high-volume data streams, temporal pattern analysis, distributional shift detection, and risk scoring.

MLOps & ML Engineering

End-to-end ML pipelines, experiment tracking with W&B, reproducible workflows, Docker, and production-quality system design.

Experience

Postdoctoral Researcher, Foundation Models & Clinical AI

Current

 Amsterdam UMC · Translational AI Lab

 2024 – Present · Amsterdam, Netherlands

  • Training transformer-based foundation models from scratch on clinical datasets with 1.2M+ records and 500+ features; architecture design, pre-training, multi-task fine-tuning, evaluation
  • Building scalable PyTorch training pipelines with distributed data loading, mixed-precision training, and W&B experiment tracking
  • Developing federated training frameworks with the University of Cambridge: multi-institutional aggregation, governance, and privacy-preserving model updates
  • Foundation models for mass spectrometry data (MassSpecFormer): pre-training on unlabelled spectra, fine-tuning for classification and retrieval
  • Publishing at npj Digital Medicine, Journal of Biomedical Informatics, and Artificial Intelligence in Medicine

Guest Researcher

 University of Amsterdam · IDEAL Lab & Complex Cyber Infrastructure Group

 2020 – 2024 · Amsterdam, Netherlands

  • Research collaboration on intelligent data engineering, privacy, and distributed computing
  • Contributing to multi-institutional projects bridging event systems and privacy-preserving analytics

Research Staff, Parrot Project

 University of Oslo · Analytical Solutions & Reasoning Group

 2020 – 2024 · Oslo, Norway

  • Multinational research initiative on event processing, data quality, and privacy in distributed systems

PhD Researcher, Real-Time Systems & LLM-Assisted CEP

 University of Groningen · Distributed Systems Group

 2020 – 2024 · Groningen, Netherlands

  • Built real-time anomaly and event detection systems using Apache Kafka and Apache Flink for high-volume IoT data streams
  • Developed GPT-CEP: LLM-assisted pipeline that automatically translates latent data patterns into interpretable detection rules deployed in live streaming environments
  • Designed privacy-preserving complex event processing: App-CEP with adaptive pattern-level privacy protection
  • Built federated learning frameworks across distributed institutional environments with runtime governance
  • Published at ACM DEBS, IEEE MSN, Elsevier Information Systems, and Springer SN Computer Science

External Member

 TU Ilmenau · Distributed Systems & Operating Systems Group

 2020 – 2024 · Ilmenau, Germany

  • Research collaboration on distributed event-based systems and quality-aware IoT processing

Education

PhD in Computer Science

University of Groningen, Netherlands

2020 – 2024

Thesis: "Adapting Event Processing in Dynamic IoT Applications: Meeting Evolving Requirements for Quality, Privacy, and Rule Autonomy"

Master's in Software Engineering

Specialization: Networks & Information Systems for IoT

Research: trade-off between quality monitoring and privacy protection in event processing systems

Bachelor's in Software Engineering

Foundation in software systems and engineering principles

Technical Skills

ML & Deep Learning

PyTorch HuggingFace scikit-learn XGBoost Transformers PEFT / LoRA Pre-training Fine-tuning imbalanced-learn Anomaly Detection

Data Engineering

Apache Kafka Apache Flink Apache Spark SQL pandas numpy Distributed Pipelines Stream Processing Time-series Analysis

MLOps & Infrastructure

Weights & Biases MLFlow Docker Git Linux CI/CD Experiment Tracking Reproducible Workflows

Programming

Python SQL C / C++ Java Bash / Shell LaTeX REST APIs

Languages

English — Fluent Dutch — Upper Intermediate Persian — Native

Projects

FBC-Transformer

Transformer-based foundation model trained from scratch on 1.2M+ clinical records. Custom tokenizer, pre-training loop with masked objectives, multi-task fine-tuning heads, and full evaluation harness with W&B tracking.

PythonPyTorchHuggingFaceW&B

MassSpectra Foundation Model

Transformer-based foundation model for high-dimensional mass spectrometry signals. Pre-training on unlabelled spectra, fine-tuning for classification and retrieval. Custom positional encodings for variable-length spectral inputs.

PythonPyTorchHuggingFace

Embedding Space Discovery

Geometry-based framework for analysing learned embedding spaces from foundation models. Clustering, label-free correspondence metrics, nearest-neighbour retrieval, and visualisation pipelines for discovering latent clinical structure.

PythonPyTorchscikit-learn

GPT-CEP

Federated LLM system for automated complex event processing rule generation and refinement. LLM reasoning combined with rule-based backends over live Kafka/Flink distributed pipelines with monitoring and feedback loops.

PythonApache KafkaApache FlinkGPT API

APP-CEP

Adaptive pattern-level privacy protection in complex event processing systems. Enforces privacy at the pattern level (not data level) using CEP-like processing in Apache Flink. Published in Elsevier Information Systems (2024).

JavaApache FlinkPrivacy

SmartScreening-CH

Clinical foundation models and ML pipelines for early detection of Congenital Hypothyroidism. Combines deep learning with classical models to support newborn screening; embedding-based patient similarity and classification.

PythonClinical AIFoundation Models

Publications

70+ citations  ·  Google Scholar, June 2026

Journal

App-CEP: Adaptive Pattern-Level Privacy Protection in Complex Event Processing Systems

Information Systems, Elsevier · 2024  · 4 citations

M. Lotfian Delouee, V. Degeler, P. Amthor, et al.

Journal

Pattern-Level Privacy Protection in Event-Based Systems

SN Computer Science, Springer · 2025

M. Lotfian Delouee, V. Degeler, P. Amthor, M. C. Schut, et al.

Conference

AQUA-CEP: Adaptive Quality-Aware Complex Event Processing in the Internet of Things

17th ACM International Conference on Distributed and Event-based Systems (DEBS) · 2023  · 16 citations

M. Lotfian Delouee, B. Koldehofe, et al.

Conference

Towards Adaptive Quality-Aware Complex Event Processing in the Internet of Things

18th IEEE International Conference on Mobility, Sensing and Networking (MSN) · 2022  · 8 citations

M. Lotfian Delouee, B. Koldehofe, et al.

Conference

Towards Federated LLM-Powered CEP Rule Generation and Refinement

18th ACM International Conference on Distributed and Event-based Systems (DEBS) · 2024  · 3 citations

M. Lotfian Delouee, D. G. Pernes, V. Degeler, et al.

Conference

Driving Towards Efficiency: Adaptive Resource-Aware Clustered Federated Learning in Vehicular Networks

22nd Mediterranean Communication and Computer Networking Conference (MedComNet) · 2024  · 7 citations

A. Khalil, M. Lotfian Delouee, V. Degeler, et al.

Workshop · Accepted

Embedding-Based Federated Learning with Runtime Governance for Iron Deficiency Prediction

FL-ICML Workshop on Federated Learning in the Age of Foundation Models (FLICS) · Accepted 2026 · with University of Cambridge

F. Zhang, S. Deltadahl, M. Lotfian Delouee, D. Kreuter, et al.

Poster

Towards Pattern-Level Privacy Protection in Distributed Complex Event Processing

17th ACM DEBS · 2023  · 3 citations

M. Lotfian Delouee, B. Koldehofe, et al.

Under Review

A Comparative Framework for Evaluating Foundation Models on Tabular Data: A Case Study in Healthcare

npj Digital Medicine

M. Lotfian Delouee et al.

Under Review

A Label-Free Correspondence Score for Validating Learned Patient Representations in Clinical Tabular Data

Journal of Biomedical Informatics

M. Lotfian Delouee et al.

Under Review

From Latent Biomarkers to Clinical Rules: Embedding-Guided Rule Mining and Attribution-Based Translation for Interpretable Tabular Learning

Artificial Intelligence in Medicine

M. Lotfian Delouee et al.

Under Review

Patient Classification via Learned Embedding Spaces: Application to Congenital Hypothyroidism Screening

CIBB 2026

M. Lotfian Delouee et al.

Under Submission

Embedding Space Discovery: Geometry-Based Clinical Data Analysis for Patient Similarity

Under Submission

M. Lotfian Delouee et al.

PhD Thesis (2025): "Adapting Event Processing in Dynamic IoT Applications: Meeting Evolving Requirements for Quality, Privacy, and Rule Autonomy", University of Groningen

News & Updates

May 2026 Talk

Presented at the Bloodcount Consortium Workshop at Cambridge's Department of Applied Mathematics & Theoretical Physics. Day 1: presented our federated learning framework for blood count data, built around the real constraints of trusted research environments, using the Flower (flwr) framework with embeddings from the DeepCBC foundation model. Day 2: joined the Foundation Model Workgroup. Ended with a dinner at Trinity College's Great Hall. Great discussions and a strong reminder of why this consortium matters.

View on LinkedIn
May 2026 Publication

Our paper on embedding-based federated learning with runtime governance for iron deficiency prediction has been accepted at the FL-ICML Workshop (FLICS) 2026, in collaboration with the University of Cambridge. Looking forward to sharing the work.

View on LinkedIn
April 2026 Research

Three papers currently under review: a comparative framework for evaluating foundation models on clinical tabular data (npj Digital Medicine), a label-free metric for validating learned patient representations (Journal of Biomedical Informatics), and a method for translating learned biomarkers into interpretable clinical rules (AI in Medicine).

View on LinkedIn

Get in Touch

Open to research collaborations, foundation model roles, applied AI positions, and data science opportunities. EU citizen, no visa required.