Open to opportunities

Med Omar Kaabar

@medomarkaabar

Message

ML Engineer building recommendation and personalization systems in production at scale.

Spain

Message

What I'm looking for

I’m looking to build and own production ML for personalization/recommendations, from feature engineering and offline evaluation to A/B testing and reliable MLOps. I want a team that values reproducibility, scalable NLP/embeddings, and measurable business impact.

I’m an ML Engineer and Data Scientist with 3+ years building recommendation and personalization systems in production. At Infor, I designed and deployed cross-sell and up-sell recommendation models using collaborative and content-based approaches, including feature engineering on large-scale transactional data, MLflow-tracked training, and FastAPI inference endpoints.

In my current role at Multiverse Computing, I build production ML pipelines for predictive maintenance, using time-series anomaly detection with MLflow experiment tracking and model versioning for reproducible deployments. I also optimize inference pipelines for large-scale models within the SingularityAI ecosystem—focusing on throughput, latency, and memory efficiency at production scale.

My NLP background spans multilingual embeddings, text classification, and LLM fine-tuning—skills I apply directly to embedding-based recommendation and item understanding at scale. Previously, at Elyadata, I benchmarked Arabic NLP models with evaluation suites and accuracy metrics, implemented document intelligence pipelines (NLP classification, NER, topic modeling), and fine-tuned multilingual models for low-resource dialects using embedding-based evaluation.

I’m driven by end-to-end model lifecycle ownership: clear offline evaluation, rigorous A/B testing design, and MLOps practices that keep experiments reproducible across generations. Even in competitions like the Mistral AI fine-tuning hackathon, I focus on full fine-tuning and evaluation lifecycles—LoRA/PEFT with HuggingFace Transformers and MLflow-tracked embedding-based metrics.

Experience

Work history, roles, and key accomplishments

Current

ML Engineer

Current

Multiverse Computing

Nov 2025 - Present (6 months)

Built predictive maintenance pipelines using time-series anomaly detection, with MLflow experiment tracking and model versioning for reproducible production deployments. Optimized large-model inference pipelines for throughput, latency, and memory efficiency within the SingularityAI ecosystem.

Anomaly Detection MLFlow Model Versioning Python fastAPI Reproducible ML

Data Scientist

Infor

Jul 2024 - Jan 2025 (6 months)

Designed and deployed cross-sell and up-sell recommendation models using feature engineering on large transactional datasets, offline evaluation, MLflow-tracked training, and FastAPI inference endpoints. Built demand forecasting and pricing optimization models with EDA, feature selection, and reproducible training pipelines, while introducing model versioning and experiment tracking for the team.

Recommendation Systems Feature Engineering MLFlow fastAPI Price Optimization Modelling

AI Engineer

Elyadata

Jul 2023 - Jul 2024 (1 year)

Developed SOTA Arabic NLP benchmarking by designing offline evaluation suites and accuracy metrics for domains without standard benchmarks. Implemented document intelligence pipelines (NLP classification, NER, topic modeling) with PostgreSQL and Elasticsearch backends and fine-tuned multilingual models for low-resource dialects using embedding-based evaluation.

Benchmarking NLP NER Topic modeling Embedding Evaluation PostgreSQL Elasticsearch

AI Research Intern

Elyadata

Feb 2023 - Jun 2023 (4 months)

Took Arabic OCR from research to production by building an Adaptive U-Net and improving precision by 30% through iterative training and custom dataset construction.

Arabic OCR Deep Learning Python

Deep Learning Intern

dB.Sense

Jul 2022 - Sep 2022 (2 months)

Built an unsupervised anomaly detection system for industrial sound using a VAE-inspired approach, achieving 80% accuracy and outperforming an auto-encoder baseline.

Anomaly Detection VAES VAE Inspired Models Deep Learning Model Comparison Python

Education

Degrees, certifications, and relevant coursework

National Engineering School of Tunis (ENIT)

Modeling for Industry & Services (MIndS)

2020 - 2023

Completed coursework focused on Machine Learning, Deep Learning, Big Data, Statistics, Optimization, Operational Research, and Linux OS under the Modeling for Industry & Services (MIndS) track.