Massyle Oumessaoud.

AI Research Scientist

Visitor, recruiter, bot or just curious ? You are welcome ! Feel free to explore and contact me :)

View Projects Get in Touch
01

About

Massyle Oumessaoud

I'm an AI Research Scientist at Uppsala University, where I design decision support tools that help medical professionals navigate uncertainty in AI-driven diagnostics. My main focus is on Bayesian classification models and their associated evaluation metrics like ROC curves (Can a 3D representation better represent uncertainty ?).

Previously, I architected and deployed a centralized data platform on Google Cloud Platform at Polynom (Paris), which completely reorganized the previous Excel-based workflow. The final product is an application that includes multiple dashboards and various data functionalities for the entire company. Before that, I published research on optimization algorithms at the University of Science and Technology of Hanoi.

I'm completing my MSc in Computer Science, AI & Data Science at UTC (First class honors). I love participating in hackathons to meet new people, discover new ideas, and take on new challenges.

Core Stack

Machine Learning

PyTorch Scikit-learn PyMC (Ongoing) MLflow Pandas / Polars NumPy Matplotlib / Seaborn / Plotly SciPy

Data Engineering

GCP Snowflake Airflow Spark SQL / NoSQL PostgreSQL Docker Kubernetes (Ongoing)

Development

Linux Python R C/C++ Git CI/CD

Generative AI

LangChain MCP Vertex AI Prompt Engineering RAG OpenClaw Computer Vision Hugging Face
0 Peer-reviewed Publication
0 Hackathon Finals
0 Production Data Platforms
0 Countries Worked In
02

Projects

AI Decision Support for Clinical Diagnostics
Featured · Research

AI Decision Support for Clinical Diagnostics

Problem: Medical staff struggle to interpret uncertainty in AI model outputs, leading to diagnostic errors.

Approach: Designed Python-based decision support tools leveraging Bayesian classification models. Evaluated performance with novel 3D ROC curve metrics and built user-friendly interfaces for non-technical clinical teams.

Impact: Reduced diagnostic uncertainty interpretation errors. Translated complex statistical outputs into actionable clinical insights.

PyTorchPandasNumPyScikit-learn
GitHub
Google AI Agents Hackathon
Hackathon Finalist

Google AI Agents — Biodiversity Monitoring

Problem: Monitoring biodiversity at scale requires processing massive satellite imagery with AI agents.

Approach: Built an MVP on Google Cloud using LangChain agents and Vertex AI to analyze satellite imagery for ecological pattern detection.

Impact: Finalist presentation before a technical jury. Demonstrated viable cloud-native AI pipeline for environmental monitoring.

LangChainVertex AIGCPSatellite APIs
GitHub
Sopra Steria Hackathon
Hackathon Finalist

Deepfake Voice Detection System

Problem: Rising deepfake audio attacks require robust voice authentication systems.

Approach: Combined ML models (SpeechBrain, Torchaudio) with rule-based expert systems to build a hybrid detection pipeline. Deployed via Streamlit for real-time testing.

Impact: Sopra Steria Hackathon finalist. Demonstrated effective deepfake detection combining neural and symbolic AI.

PyTorchSpeechBrainTorchaudioStreamlit
Centralized Data Platform
Product

Centralized Data Platform — Polynom

Problem: Fragmented client data sources causing slow processing, high costs, and data inconsistency.

Approach: Architected a unified data platform on GCP with ELT pipelines (Airflow, FastAPI, PostgreSQL), Keycloak for auth, and Grafana for monitoring.

Impact: Cut processing time significantly. Enabled sales teams to access unified analytics without engineering support.

GCPFastAPIAirflowPostgreSQLDocker
Healthcare Data Warehouse
Data Engineering

Healthcare Data Warehouse

Problem: Sensitive medical data required secure, traceable ETL pipelines compliant with strict healthcare standards.

Approach: Engineered end-to-end ETL pipelines on Snowflake with Spark, implementing full audit trails and CI/CD for reproducibility.

Impact: Achieved full traceability compliance for medical data processing workflows.

SnowflakeSparkSQLCI/CD
GitHub
03

Research & Publications

Peer-Reviewed Paper

Optimal Design of a Magnetic Sensor for a Linear Machine

EEE-AM 2023 · IEEE · Co-authored

Optimized magnetic sensor designs using finite element analysis and particle swarm optimization, maximizing detection performance while respecting peak flux density constraints.

DOI: 10.1109/EEE-AM58328.2023.10394938
2023
Ongoing Research

3D ROC Curves for Bayesian Classification Evaluation

Uppsala University · 2025–Present

Developing novel evaluation metrics for Bayesian classification models using three-dimensional ROC surfaces. Building decision support interfaces for clinical teams.

2025
04

Experience

Dec 2025 — Present

AI Research Scientist

Uppsala University · Stockholm, Sweden

  • Designed decision support tools for medical staff to interpret AI model uncertainty
  • Evaluated Bayesian models with novel 3D ROC metrics and built user-friendly interfaces for clinical teams
PyTorch · Pandas · NumPy · Scikit-learn
Aug 2024 — Mar 2025

Data Engineer & Data Scientist

Polynom · Paris, France

  • Architected a centralized data platform unifying client data sources
  • Developed ELT pipelines translating business needs into technical solutions
  • Prototyped AI tools to automate internal processes
GCP · FastAPI · PostgreSQL · Airflow · Docker
Apr 2023 — Jul 2023

Research Assistant

University of Science and Technology of Hanoi · Vietnam

  • Optimized magnetic sensor designs using finite element analysis and PSO
  • Co-authored IEEE peer-reviewed paper on sensor optimization
Python · Ansys · MATLAB
2023 — 2026

MSc Computer Science, AI & Data Science

Université de Technologie de Compiègne (UTC) · France

  • Excellence with Honors — Data Warehousing, Computer Vision, Algorithms & Optimization
2020 — 2023

Bachelor's in Electrical Engineering (Top of the class)

CY Cergy Paris Université · France

05

Technical Notes

Coming Soon

Articles on electronics, applied ML, and lessons learned from real-world projects are on the way.

06

Get in Touch

Whether you're interested in research collaboration, have a data challenge to solve, or just want to discuss the future of AI — I'd love to hear from you.