Hello, I'm

Ali AbuSaleh

Senior ML & Generative AI Engineer

ML Engineer at Goethe University's Text Technology Lab, Frankfurt. I build production LLM agents, multimodal AI systems, and scalable ML infrastructure for 6 research teams across 5 institutes — with 4 peer-reviewed publications across ACM, LREC, ICNLP, and Diabetes Technology & Therapeutics.

LLM Agents Multimodal AI Healthcare AI Arabic NLP
  View Research   Let's Collaborate
Ali AbuSaleh
Rank 2/15 AraSentEval
4 Publications
6 Research Teams

About Me

Senior AI Engineer and Researcher — here's what I bring to the table

Who I am

Senior AI Engineer with 5+ years building production Generative AI, LLM agents, and multimodal systems. At Goethe University's Text Technology Lab (Feb 2025 – present) I lead open-source AI infrastructure used by 6 research teams across 5 institutes — ingestion pipelines, vector stores, and model-serving, all open-source.

I hold an Erasmus Mundus M.Sc. (EU Excellence Scholarship, A+ / 16.4/20) in Big Data Management & Analytics from ULB Brussels, Univ. Paris-Saclay, and UPC Barcelona. Before academia I spent 2+ years as a backend engineer at Harri Technologies (AWS / Elasticsearch) and Exalt Technologies (Spring Boot / C++).

5+
Years
experience
4
Peer-reviewed
publications
6
Research
teams led
2B+
Records
processed

How I can help

Research Collaboration

Joint papers, co-authorship, and shared experiments in multimodal AI, NLP, or healthcare AI. Comfortable with arXiv preprints and peer-reviewed venues.

GenAI & ML Engineering

LLM agents (LangChain, CrewAI, LlamaIndex), RAG pipelines, multimodal systems, and cloud deployments on AWS — production quality.

PhD Mentoring

Thesis guidance, paper writing support, and research methodology for students working in ML, NLP, or multimodal AI.

Research

4 peer-reviewed publications (2025–2026), 3 more in preparation — spanning multimodal AI, NLP, and healthcare AI. Google Scholar ↗

Multimodal AI

Video + audio + text fusion, tri-modal topic modeling (MMTM), cross-modal negation detection — +7.03% F1 over unimodal baselines using VLMs + JEPA2.

Healthcare AI

Continuous glucose monitoring (CGM) foundation models, 3-hour blood glucose forecasting (6× prior state-of-the-art), multitask VAE — 2B+ records processed.

Arabic NLP

Multi-dialectal Arabic sentiment analysis (SARF, Rank 2/15 internationally, Macro-F1 0.9263), parliamentary corpora (MultiParTweet), VLM annotation.

Agents & GenAI

Open-source LLM orchestration infrastructure for 6 research teams across 5 institutes. Multi-agent research writer (CrewAI), agentic RAG, tool-use agents.

New ACM Web Science 2026 2026

From Images to Topics: Evaluating Vision-Language Models for Topic Classification of Election Advertising

Weiss, Abusaleh et al. — VLM + JEPA2 framework for cross-modal negation; +7.03% F1 over unimodal baselines.

View on Scholar
New ICNLP 2026 2026

Learning to Detect Cross-Modal Negation: An Analysis of Latent Representations and an Attention-Based Solution

Abusaleh et al. — Accepted at ICNLP 2026, Xi'an, China.

View on Scholar
New LREC-COLING 2026 2026

TTLab at AraSentEval: SARF — Sentiment Analysis via Root-based Fusion for Multi-Dialectal Arabic

Abusaleh et al. — OSACT7 @ LREC-COLING 2026. Rank 2/15 internationally, Macro-F1 0.9263 (BERT-CNN-BiLSTM-Attention).

View on TTL
Diabetes T&T 2025 2025

Generative AI on CGM: Towards a Foundation Model for Glucose Prediction, Root Cause Analysis and Anomaly Detection

Rahim & Abusaleh — Published in Diabetes Technology & Therapeutics, Vol. 27, 2025. 3-hour forecasting, a 6× extension of prior 30-min state-of-the-art.

Read paper
  Google Scholar   TTL Lab Profile

Updates

Recent news — paper acceptances, preprints, and positions

Jun 2026 Paper

Paper accepted at ACM Web Science Conference 2026: "From Images to Topics: Evaluating Vision-Language Models for Topic Classification of Election Advertising"

View on Scholar
Jun 2026 Paper

Paper accepted at ICNLP 2026: "Learning to Detect Cross-Modal Negation: An Analysis of Latent Representations and an Attention-Based Solution"

View on Scholar
May 2026 Preprint

New arXiv preprint: "MMTM: Tri-Modal Topic Modeling for Long-Form Video via Similarity-Gated Fusion"

View preprint
Apr 2026 Paper

Paper accepted at LREC-COLING 2026 (OSACT7): "TTLab at AraSentEval: SARF — Sentiment Analysis via Root-based Fusion for Multi-Dialectal Arabic" — Rank 2/15 internationally

View on TTL
Feb 2025 Position

Joined Text Technology Lab, Goethe University Frankfurt as ML Engineer — leading open-source AI infrastructure for 6 research teams across 5 institutes.

TTL Lab profile
Mar 2025 Paper

Published in Diabetes Technology & Therapeutics (Vol. 27): "Generative AI on CGM: Towards a Foundation Model for Glucose Prediction, Root Cause Analysis and Anomaly Detection"

Read paper
Oct 2024 Preprint

arXiv preprint: "A Multitask VAE for Time Series Preprocessing and Prediction of Blood Glucose Level" (arXiv:2410.00015)

Read on arXiv

Writing

Technical articles and notes

Multiprocessing in PyTorch
July 3, 2024 PyTorch Multiprocessing

Leveraging Multiprocessing in PyTorch

How to use torch.multiprocessing to significantly boost performance when handling large-scale data and parallel processing workloads.

Read on Medium

Get In Touch

Available for research collaboration, GenAI & ML engineering, and PhD mentoring

Contact Information

Location

Frankfurt am Main, Germany

Email

[email protected]

Work Authorization

EU Blue Card — full right to work in Germany