JEREMY JOUVANCE

As a Technical Leader, I am responsible for several innovative projects. My duties include managing a team of eleven, recruiting new talent, and liaising with stakeholders. I also oversee architectural choices, quality control, and the implementation of solutions.

• Pioneered the initial AI experiments at Amundi, navigating significant technical challenges including strict security policies.
• Creator of Alto Studio to respond the lack of A.I tools

  • Studio Data

    • Architected and developed an on-premise, scalable "All Data Platform" using lakehouse architecture.
    • Ensured compliance with financial data retention regulations.

  • Studio Business Intelligence

    • Facilitated BI capabilities via Trino and Superset.

  • Studio Coding

    • Created an in-browser secured coding workspace for citizen developers.

  • Studio A.I

    • Developed a Secure Chat solution with a microservices architecture (Django/Postgres/React) for seamless LLM integrations.
    • Built an LLM gateway to simplify model switching and control access/consumption across business cases.
    • Implemented a hybrid Rag system (vector search, BM25, knowledge graph) for document analysis across multiple data sources.
    • Developed multi-agent systems for task solving.
    • Delivered multiple financial use cases, including:
      • Deep Research for Investment Specialists
      • Studio Bots for creating LLM agents directly in the browser
      • Podcast generator for market review
      • And more...

Ideolys develops back-office software for large restaurant chains. A key challenge addressed was the discrepancy between ordered and delivered quantities, which affects pricing and necessitates adjustments.

• Automated integration of food bills via phone scan (OCR) using Python, Tesseract, Tabula, and OpenCV.
• Implemented automatic detection of quantities and prices for bill adjustment.
• Developed replacement and adjustment of recipes via nearest neighbor techniques (NLTK, spaCy).
• Integrated, developed, and tested VueJS components (Vue.js, RabbitMQ, Flask).

Emoleaks marked the completion of my academic journey and introduced a revolutionary approach to advertising testing by utilizing real-time emotional analysis.

• Sourced and curated a diverse dataset to train a CNN, ensuring robust model performance across varied demographics.
• Employed data augmentation techniques to expand the dataset, enhancing model accuracy and reliability.
• Developed and implemented a Convolutional Neural Network (CNN) model for real-time detection and analysis of viewers’ emotions during media exposure using TensorFlow, Keras, and Flask.