Aissam Djahnine
CV
2× winner, JFR radiology data challenge
Medical imaging · Computer vision · Applied AI

AI Research Engineer @ DentalMonitoring

PhD in Computer Science with 5+ years building computer vision for healthcare, from pathology detection in 3D CT scans to remote orthodontic monitoring at DentalMonitoring.

View selected work Download CV Email me
GitHub LinkedIn Google Scholar ORCID ResearchGate
Core stack Python PyTorch TensorFlow Docker MLflow GitLab Git GitHub
Portrait of Aissam Djahnine
5+
years in medical AI
6
papers and preprints
64
citations · h-index 3
Top 6%
DaT SPECT challenge · 59th of 1,009

Experience

Full CV →

AI Research Engineer @ DentalMonitoring

October 2024 – Present
Supervisor: Guillaume Ghyselinck

I design and deploy the computer vision systems that power remote orthodontic monitoring, transforming raw patient data into reliable clinical insights for doctors worldwide.

PhD Candidate | Machine Learning Engineer @ Philips Health Technology Innovation – AI Research Hub France

July 2021 – September 2024
Supervisors: Pr. Loic Boussel, Dr. Nicolas Villain
  • Researched 3D pathology detection in CT on a 40k-scan multi-hospital database: contrastive self-supervised pretraining (SimCLR) raised 4-pathology F1 from 0.79 (ImageNet init) to 0.94, with 3D localization of findings.
  • Developed CT-3DFlow, an unsupervised, position-conditioned 3D normalizing flow trained on healthy scans only that produces 3D anomaly maps and outperformed 8 anomaly-detection methods.
  • Published 5 peer-reviewed papers (3 first-author), including in Medical Physics and Diagn. Interv. Imaging.
  • Won 1st place twice at the French Society of Radiology AI challenge with the Philips–HCL team (JFR 2022 pulmonary embolism, JFR 2023 pancreatic mass), with 40+ teams across both editions, including GE HealthCare and Guerbet.

Computer Vision and Deep Learning Intern @ GE Healthcare

March 2020 – August 2020
Supervisors: Pr. Serge Muller, Dr. Andrei Petrovskii

Investigation of automatic search methods for neural network hyperparameters (Neural Architecture Search). Application to Mammographic Data within WHARe team (Women's Health Applied Research)

  • Extended neural architecture search (PC-DARTS) with a cell-initialization constraint and applied it to 5-class mammography classification (DDSM), beating the research team's hand-designed model (78.1% vs 75.0% accuracy).

Education

July 2021 – September 2024

Université Claude Bernard Lyon 1 – CREATIS Lab (MYRIAD team)

PhD in Computer Vision & Medical AI · Lyon, France

Thesis: Automated Pathology Detection in Conventional CT Scans. CIFRE industrial doctorate with Philips Health Technology Innovation France and HCL.

2018 – 2020

Sorbonne University & Télécom Paris

Master’s Degree in Computer Science, IMA track (Images, Computer Vision & Graphics) · Paris, France · Grade: 14.66/20

Master 2: Image Processing & Computer Vision (IMA). Pattern Recognition for Image Analysis and Interpretation, Advanced Image Processing, Advanced Computer Vision Techniques, Seminars and practice IMAGES, Advanced Machine Learning and Deep Learning.

Master 1: Engineering of Robotics and Intelligent Systems (ISI). Random Signal Processing, Statistics, Embedded Systems & Sensors, Artificial Intelligence, Computer Vision, Robotics, Image Processing.

September 2017 – August 2018

Pierre and Marie Curie University (Sorbonne Université)

Bachelor’s Degree, Electrical and Electronics Engineering · Paris, France

Signals and systems, Magnetic induction and energy conversion, Mathematical tools for electronics, Programming and numerical methods, Smart grids and energy management, Techniques and devices for analog electronics, Digital systems and embedded processors, Radiation and propagation, Robotic systems.

September 2016 – August 2017

Ecole Nationale Polytechnique d'Oran (ENPO)

Engineer’s Degree, Electrical and Electronics Engineering · Oran, Algeria

Major: Electrical Engineering. Analog electronics, Digital systems, Electric and magnetic circuits, systems theory, Continuous linear systems, Measurement techniques, Power electronics, Sampled servo systems, State-space analysis and control, Instrumentation, Renewable energies and sustainable development.

September 2014 – August 2016

Ecole Préparatoire aux Sciences & Techniques (EPST)

Preparatory Classes for Engineering Schools · Annaba, Algeria

Studied advanced Mathematics, Physics, Chemistry, and Engineering Sciences. Prepared for the entrance exams to top engineering schools in France.

Honors and awards

Philips team receiving first place at the JFR 2023 AI challenge
2023 · Winner

JFR 2023 Data Challenge WINNER: Pancreatic masses detection in 3D CT scans

I was a member of the Philips team in collaboration with Hospices Civils de Lyon that won the JFR (les Journées Francophones de Radiologie) data challenge. The solution used a deep learning-based algorithm to detect pancreatic masses in 3D CT scans.

Announcement ↗ Challenge paper ↗
Philips team celebrating first place at the JFR 2022 AI challenge
2022 · Winner

JFR 2022 Data Challenge WINNER: Pulmonary embolism detection in 3D CT scans

I was a member of the Philips team in collaboration with Hospices Civils de Lyon that won the JFR (les Journées Francophones de Radiologie) data challenge. The solution used deep learning for computer vision to detect pulmonary embolism in CT scans.

Announcement ↗ Challenge paper ↗ Our Solution (Journal Paper) ↗

Research

Publications →
PhD thesis · Université Claude Bernard Lyon 1

Automatic Detection of Pathologies in Conventional Scanner Data

This thesis investigates AI-based detection and characterization of pulmonary pathologies in conventional CT scans, with a focus on reducing annotation requirements. It explores supervised, weakly supervised, and unsupervised learning strategies to detect anomalies and support radiologists in identifying incidental findings beyond the initial clinical indication.

Supervisor: Loïc Boussel · Lab: CREATIS · Field: Biomedical imaging
Thesis overview → Read manuscript ↗ HAL ↗ theses.fr ↗
Defended
Sep 17, 2024
Views
152
Downloads
54
HAL stats as of Oct 19, 2026
Featured project

Segmentation and Measurement of Skeletal Muscle Areas on CT Scans

Developed a U-Net that segments and quantifies skeletal muscle in CT at the L1, L2 and L3 vertebral levels for sarcopenia assessment, now used as a clinical assistance tool at HCL.

Medical imagingSegmentation · CT
View project →
Normal and abnormal DaT SPECT brain scans shown in axial, coronal and sagittal views
Challenge · Parkinson's disease

DaT SPECT Abnormality Classification

Built an end-to-end pipeline to classify DaT SPECT scans as normal or abnormal, combining four CNN families with striatal binding-ratio features. The solution achieved an AUROC of 0.9379 and ranked 59th on the private leaderboard; 1,009 participants joined the competition.

0.9379
AUROC
59th
of 1,009 participants
Medical imaging3D CNN ensemble
Project page → GitHub ↗ Release v1.0 ↗

CT-3DFlow — Unsupervised Pulmonary CT Analysis

Learning the normal appearance of 3D pulmonary CT scans with normalizing flows to detect abnormalities.

arXiv2024

eCAPRI: a novel tool combining clinical and imaging data for post-TAVI mortality prediction

eCAPRI combines clinical information with automatically extracted CT biomarkers to estimate one-year mortality after transcatheter aortic valve implantation (TAVI).

European Radiology2025

Weakly-Supervised learning based Pathology Detection and Localization in 3D Chest Scans

This study identifies multiple chest abnormalities and highlights their locations in 3D CT scans, combining patch-level predictions into a patient-level assessment.

Medical Physics2024

Detection and severity quantification of pulmonary embolism with 3D CT data using an automated deep learning-based artificial solution

This work investigates automated pulmonary embolism detection and severity assessment from three-dimensional CT pulmonary angiography examinations with limited annotations.

Diagn. Interv. Imaging2024

BERT-based natural language processing analysis of French CT reports: Application to the measurement of the positivity rate for pulmonary embolism

This study uses natural language processing to analyze French CT reports and measure the diagnostic yield of examinations requested for suspected pulmonary embolism.

ReDII2023

Tailored 3D CT contrastive pretraining to improve pulmonary pathology classification

This work explores contrastive pretraining for pulmonary pathology classification using three-dimensional chest CT patches.

IEEE ICSP2022

Other research projects

Publications

Google Scholar →
Cited byAllSince '21
Citations6464
h-index33
i10-index22
Google Scholar, as of Nov 11, 2026

Projects and open source

All projects →
Open source · Contributor · 2026

llmfit

Hundreds of models & providers. One command to find what LLM runs on your hardware.

My pull requests ↗ Repository ↗
  • Added model comparison workflow (llmfit diff) and TUI compare mode.
  • --sort ranks fit results quickly for easier side-by-side comparison.
  • MoE token/s estimation now uses active parameters, with clearer baseline speed labels.
  • --max-context adds a safe upper bound for memory estimation on limited hardware.
v0.6.2 v0.5.6 v0.5.5 v0.4.9

Featured projects

Normal and reduction cells discovered by DARTS

Neural Architecture Search for mammography (DARTS)

Master's thesis at GE Healthcare: DARTS architecture search for mammogram classification. The searched architecture beat the hand-designed baseline, 78.05% vs 75.03% accuracy on DDSM-Patches.

GE Healthcare · 2020AutoMLROC-AUC 0.92
Project page →Internship details →
Frame and sequence GAN architecture for video inpainting

Unsupervised Spatiotemporal Data Inpainting

Course project with Ilyas Aroui and Abdelraouf Keskes: reimplementing unsupervised video inpainting with frame and sequence GAN discriminators on SST, FaceForensics++, KTH and BAIR.

GANVideoSorbonne · 2020
Project page →GitHub ↗
Scanned museum label with detected text lines boxed in blue

Handwritten Text Recognition System

Transcribing scanned handwritten museum labels: OpenCV line detection, Canny or YOLO character detection, and a Keras classifier for digits and letters (97% test accuracy).

OCRComputer vision2021
Project page →

Conferences and summer schools

16th IEEE International Conference on Signal Processing (ICSP)

21 – 24 October 2022 · Beijing, China

Presented Tailored 3D CT contrastive pretraining to improve pulmonary pathology classification.

Oxford Machine Learning Summer School (OxML 2023)

13 – 16 July 2023 · Oxford, United Kingdom

ML x Health track, in person. Certificate ↗

Resume

Download PDF →

Internships

Computer Vision and Deep Learning Intern @ GE Healthcare

March 2020 – August 2020Buc, France
Supervisors: Pr. Serge Muller, Dr. Andrei Petrovskii

Investigation of automatic search methods for neural network hyperparameters (Neural Architecture Search). Application to Mammographic Data within WHARe team (Women's Health Applied Research)

  • Extended neural architecture search (PC-DARTS) with a cell-initialization constraint and applied it to 5-class mammography classification (DDSM), beating the research team's hand-designed model (78.1% vs 75.0% accuracy).

BIOMEDInnov Exchange Program Intern @ University of Waterloo

May 2019 – August 2019Ontario, Canada

Selected from my Electrical Engineering Master’s program for the BIOMEDInnov exchange at the University of Waterloo, a venture aimed at pioneering Biomedical Engineering solutions. Contributed to a project designing energy-harvesting sensors and wireless systems to enhance sustainable biomedical technologies. The program involves Master students in research projects and strengthens innovation and entrepreneurship in biomedical engineering.

Program detail: Developing sensors and wireless systems that can harness energy from the surrounding environment.

Skills: Electronics · Antennas · Wireless Sensor Networks · Smart Antennas · Energy Efficiency · Linear Algebra

Integrated Sensor Diagnostics Intern @ Halliburton

March 2017 – April 2017Hassi Messaoud, Ouargla, Algeria

Observational internship at Halliburton Energy Services (Algeria Division) focused on integrated sensor diagnostics. Gained insights into deployment and maintenance of advanced sensor systems used in oilfield services, covering data acquisition, signal processing, and diagnostics. Shadowed engineers in calibration and troubleshooting of sensor equipment, contributing to data integrity and operational efficiency.

Skills: Sensors

Engineering Intern @ Sonatrach

December 2016 – January 2017Hassi Messaoud, Algeria

Observational internship in the maintenance department of the oil industry. Gained understanding of operational challenges and maintenance strategies critical to energy production, with exposure to methods for sustaining and improving large-scale infrastructure.

Skills: Engineering · Knowledge Engineering

Academic services

2023Reviewer

20th IEEE International Symposium on Biomedical Imaging (ISBI 2023)

Reviewed 5 articles for publication focused on AI and medical imaging (CT, MRI, X-ray).

October 2018 – April 2019Paris, France

Academic Tutor @ Pierre and Marie Curie University (Sorbonne Université)

Fulfilled the role of an Academic Tutor, guiding undergraduates (L1-L3) through academic projects and lab sessions. Specialized in Mathematics and Linear Algebra exercises, and provided hands-on mentorship in Analog and Digital Electronics. Supported programming proficiency in C, MATLAB, and VHDL.

Skills: MATLAB · VHDL · Linear Algebra · C (Programming Language) · Teaching

Certifications

Python 3Codingame

Certification Codingame - Python 3

Credential ID: Wch0EvJDnkCgVQlsGaB7vQ

Show credential ↗
July 2023AI for Global Goals

Oxford Machine Learning Summer School (OxML 2023) - ML x Health Track, In Person

Skills: Deep Learning · Machine Learning · Artificial Intelligence (AI)

Show credential ↗
April 2020Coursera

Structuring Machine Learning Projects

Credential ID: 861d698b06df950442c51cd814fc282c

Show credential ↗

Beyond the lab

Running

1 half marathon · 3 × 10 km
Milano21 2025 result
21 km

Milano21 Half Marathon

Milan · 2025 · Results →
Odyssea Paris 2025 result
10 km

Odyssea Paris

Paris · 2025 · Results →
Run for Planet medal
10 km

Run for Planet

2025 · Event page →
10 km de Parilly 2024 result
10 km

10 km de Parilly

Lyon · 2024 · Results →

Travel

21 countries · 3 continents
VisitedNot yetDrag to pan · use the buttons to zoom