Skip to main content

Bilel Saghrouchni

📍 Lyon | bilel.saghrouchni@gmail.com | linkedin.com/in/bilel-saghrouchni | bilelsgh.github.io | 🇫🇷: Native 🇬🇧: Fluent

Profile​

Applied AI Engineer and PhD Candidate with over 4 years of experience at the intersection of cutting-edge research and enterprise deployment. Specialized in designing autonomous agentic systems (Deep Reinforcement Learning, LLMs) and translating complex business requirements into robust, scalable AI solutions.

Experience​

PhD Candidate in Machine Learning / Applied AI Engineer — INSA Lyon / SPIE ICS​

🗓️ Nov 2023 – Present

  • Enterprise AI Deployments: Direct collaboration with hospital stakeholders to deploy autonomous supervision systems, translating complex clinical requirements into robust technical artifacts.
  • Autonomous Agent Design: Design and development of RL agents (PPO, A2C, Double DQN) operating in fully unsupervised and semi-supervised environments, with novel reward mechanisms based on ensemble learning and fuzzy-clustering signals.
  • End-to-End ML Pipelines: Data ingestion, feature extraction, representation learning (Autoencoders, Contrastive Learning) and evaluation — MLflow integration for experiment tracking and reproducibility.
  • Production Infrastructure: Deployment of high-availability inference services via Dockerized REST APIs in constrained hospital IT environments.
  • Applied Research: Continual learning, online adaptation and dynamic policy optimization; 3 publications (KES 2025, CSNet 2024, KES 2026).
  • Mentoring: Supervision of engineering interns and conduct of code reviews.

Stack: Python, PyTorch, TensorFlow, Gymnasium, Stable-Baselines3, MLflow, Docker, PostgreSQL

Software Engineer — Devoteam​

🗓️ June 2022 – September 2023 | Levallois-Perret, France

  • Technical Consulting: Served as a technical consultant for enterprise clients, identifying business challenges and delivering tailored software solutions on the ServiceNow platform.
  • Conversational AI: Development and production deployment of a voice assistant for field technicians, integrating NLP pipelines, speech recognition and task automation.
  • Federated Learning: Implementation of a privacy-preserving federated learning system for distributed model training in constrained client environments.

Stack: Python, TensorFlow, Flask, Rasa, VOSK

Full-Stack Software Engineer — ETIC (INSA Lyon)​

🗓️ Sept 2021 – April 2023 | Lyon, France

  • Design and deployment of a high-availability full-stack application for a national transportation company.
  • Development of high-throughput backend pipelines and cloud infrastructure management (AWS Elastic Beanstalk).

Stack: Angular, Flask, PostgreSQL, AWS Beanstalk

Data Scientist (Intern) — Worldline​

🗓️ April – August 2021 | Villeurbanne, France

  • Processing of large financial datasets for training fraud detection models.
  • Development of data pipelines and interpretability dashboards for business stakeholders.

Stack: Pandas, Scikit-learn, SQL

Tech Stack​

PythonTensorFlowGymnasiumStable-baseline3Scikit LearnLangchainStreamlitML-FlowUVpre-commit

Education​

INSA Lyon — Engineering Degree, Computer Science & Telecommunications | 2019 – 2022 Top 5% of cohort. Courses: Machine Learning, Deep Learning, Reinforcement Learning, Distributed Systems.

National University of Singapore (NUS) — Exchange Program, School of Computing | 2022 Specialization: Computer Security, Feature Engineering, Information Retrieval.

Research & Publications​

  • Comprehensive Comparison of Streaming Clustering-Based IDS — KES 2025 Bilel Saghrouchni, Frédéric Le-Mouël, Bogdan Szanto [DOI]
  • Towards An Unsupervised Reward Function For A Deep RL Based IDS — CSNet 2024 Bilel Saghrouchni, Frédéric Le-Mouël, Bogdan Szanto [DOI]
  • Deep Reinforcement Learning for Continuous Intrusion Detection Using Unsupervised Fuzzy Clustering-Based Rewards — KES 2026 Bilel Saghrouchni, Frédéric Le-Mouël, Bogdan Szanto

AI & LLM Projects​

ask_zotero — RAG on Academic Library Production-ready RAG system for querying and synthesizing scientific literature. Vector embeddings, LLM retrieval pipelines, advanced prompt engineering. (LangChain, Streamlit)

ExplainMe — LLM Content Generator Automated LLM tool for generating explanations and summaries. Agentic workflow chaining and output reliability evaluation.

Data Explorer — Interactive ML Tool Interactive dataset exploration and preprocessing interface for ML engineers.

Skills​

AI & Machine Learning Deep Learning, Reinforcement Learning (PPO, A2C, DQN), Representation Learning, Agentic Systems, Continual Learning, Online Learning

LLM & Generative AI RAG, Prompt Engineering, LLM Applications, Agent Orchestration — Active interest in: SFT, RLHF, PEFT, Synthetic Data Generation

Engineering & Deployment Python (expert), PyTorch, TensorFlow, Scikit-learn, Docker, REST APIs, MLflow, PostgreSQL, AWS

Languages French (native) — English (fluent)

Awards​

  • 🥇 1st place (national) — Devoteam Devogame (AI & Innovation, 2021)
  • 🏅 3rd place — Hackathome by Accenture (combinatorial optimization, 2022)
  • 🏅 3rd place — Safran Black Out Challenge

Interests​

  • AI outreach & content creation (podcast, blog)
  • Judo — Black belt, 18 years of practice, national competitor, referee
  • Photography

📄 Download the CV

Last update: May 2026