Quentin Possamaï

Data Scientist - MLOps | Alstom

Switzerland

About

Currently Data Scientist at Alstom, specializing in machine learning for industrial applications. Worked on a PhD focused on Deep Learning and Control for Stable Drone Flight. Holds a Master’s degree in engineering with a data science major from École Centrale de Lille.

Experience

  • Data Scientist at Alstom
    Nov 2022 - Present · 3 yrs 8 mos

    Anomaly detection in railway logs - Combined log parsing and Large Language Models (LLM), prompt engineering, and log clustering (Drain, Spell, Brain, BerTopic, LangChain, PyTorch). Built a domain knowledge database (PostgreSQL) to leverage in-context learning to detect anomalies. Web app deployment - Developed and deployed a full-stack web app using Streamlit, Grafana, Docker, Kubernetes, Flask APIs, PostgreSQL, and Elasticsearch for data collection and anomaly labeling.

  • Doctorant Laboratoire LIRIS - Deep Learning and Control Theory for Stable Drone Control at LIRIS
    Jan 2020 - Nov 2022 · 2 yrs 11 mos

    Self-supervised learning - Predicted drone turbulence using depth cameras with CNNs, GRUs, MLPs, depth map auto-encoding, and sim-to-real evaluation. Diverse data sources - Worked with time series, RGB-D, stereo images, and simulations from both real and simulated dynamic systems. Drone control - System identification, dynamic system modeling, and model predictive control. Drone simulation - Simulated drone flights and air perturbations in Habitat with Matterport3D scanned environments.

  • Machine Learning Intern at Accenture
    Apr 2019 - Sep 2019 · 6 mos

    Simulation development - Created a supply chain environment for autonomous agents. Deep reinforcement learning - Applied Deep Q-Learning and Deep Deterministic Policy Gradient to optimize agent behavior.

  • Data Scientist Intern at Enedis (ex ERDF)
    May 2018 - Sep 2018 · 5 mos

    Time-series analysis - Analyzed smart grid data for the deployment of electric smart meters in France. Machine learning - Applied regression algorithms (Random Forest, SVM, Logistic Regression) to predict power spikes from energy consumption data.

  • Restaurant Assistant at オーパヴィラージュ (Opa Village)
    Jul 2017 - Aug 2017 · 2 mos