PDF

DATA ❦ GAZETTE

Jules Lascoux

Data Analyst
Jules Lascoux
Fig. 1 - Jules Lascoux

I am a passionate data analyst, a bit of a generalist, I am able to approach cross-cutting topics. I enjoy discovering what I don't know.

My work focuses on data analysis, I make everything I can visual and exploitable. I master Python, SQL and PowerBI. My curiosity has allowed me to explore many domains related closely or distantly to data

Location: Toulouse, France
Email: jlslascoux@gmail.com
LinkedIn: linkedin.com/in/jules-lascoux/
GitHub: github.com/JulesLscx
  • Data Analysis: Pandas, DuckDB, Excel
  • Machine Learning: Scikit-learn, TensorFlow, PyTorch, XGBoost
  • Visualization: PowerBI, Matplotlib, Seaborn
  • Languages: Python, SQL, Java
  • Tools: Oracle, AWS, Docker, Git

⚘ Professional Experience ⚘

◆ Data Analyst

Banque Populaire Occitane Toulouse, France
Sep 2024 - Present

Developing data analysis tools for a logistics service

  • Created tools for analyzing electrical consumption
  • Tracked and analyzed site failures and interventions
  • Built interactive dashboards with PowerBI
  • Automated data pipelines with Power Automate

◆ Software Developer

Komax France Toulouse, France
Apr 2023 - Sep 2024

Developing software modules for importing electrical design data from third-party software

  • XML data parsing with Java
  • Object-oriented design of data classes
  • Optimization of processing (reduced 25% write times)
  • Writing technical documentation

⚘ Education & Diplomas ⚘

◆ Master in Miage

Université Toulouse Capitole Toulouse, France
Sep 2024 - Sep 2026

Relevant courses: Big Data, Data Analysis, AI

◆ Bachelor in Computer Science

Université Paul Sabatier Toulouse, France
Sep 2021 - Jul 2024

Relevant courses: Database Architectures, Machine Learning, Object-Oriented Programming

⚘ Featured Projects ⚘

CLASSIFIED ADS & PROJECTS PAGE 1 OF 2

Implementation of DS-STAR

Agent framework for Data Science automation

PythonLLMGeminiMulti-Agent
Click for details ↻
Implementation of DS-STAR

Implementation of Google's research paper "DS-STAR: A State-of-the-Art Versatile Data Science Agent". This framework uses a multi-agent system powered by Gemini models to analyze data, plan, write and execute code to answer user queries.

Key features:

  • Agentic Workflow: Pipeline of specialized agents (Analyzer, Planner, Coder, Verifier, Router, Debugger, Finalyzer).
  • Reproducibility: Complete saving of steps, prompts, code and results for auditability.
  • Interactive & Resumable: Ability to pause, resume and execute step by step.
  • Editing & Debugging: Allows manual code editing and includes an automatic debugging agent.

Implementation of ProFiT

Program search for algorithmic trading via LLM

PythonLLMEvolutionary Algorithms
Click for details ↻
Implementation of ProFiT

Reproduction of the ProFiT framework (Program Search for Financial Trading). It is an LLM-driven evolutionary search algorithm that autonomously discovers and improves algorithmic trading strategies.

Key features:

  • LLM-driven Evolution: Uses LLMs (e.g., GPT-4) to act as "Analyst" and "Developer" to mutate strategies.
  • Code-level Search: Strategies are represented as executable Python code.
  • Self-Repair: Includes a repair loop where the LLM fixes syntax or execution errors.
  • Walk-Forward Validation: Implements cross-validation specific to time series.
CLASSIFIED ADS & PROJECTS PAGE 2 OF 2

Series-Search

Python API to search for TV series via their subtitles (Original/French)

PythonElasticsearchAPINLP
Click for details ↻
Series-Search

Development of a search engine that finds TV series from their subtitles. The project indexes 128 series and allows searches in original language (VO) and French version (VF).

Key features:

  • Advanced text search: Using Elasticsearch to index and search subtitles.
  • Multilingual support: Handles subtitles in VO and VF.
  • REST API: Interface to interact with the search engine.
  • Database: Indexing of 128 popular series.

RankFinder

Estimating ratings from customer comments via Deep Learning

PythonJupyter NotebookDeep LearningNLP
Click for details ↻
RankFinder

University project aimed at evaluating customer review polarity (Amazon, Spotify, etc.) to predict the associated rating. The objective is to help platforms better understand consumer feedback.

Key features:

  • Sentiment Analysis: Evaluation of the polarity of text comments.
  • Rating Prediction: Estimation of the score (rating) based on review text.
  • Machine Learning / Deep Learning: Use of learning models for natural language processing.
}