.png)
Titouan Verhille
Currently learning more about alignment
Overview
Student in Artificial Intelligence / Natural Language Processing @EPITA
Technical Lead @Datapix
Recent projects: Mistral Hackathon MCP Finalist, TimeGuessr LLM Benchmark and Project for French Ministry for Armed Forces — see all projects.
Paris, France
Experience
Datapix Agency
Current Employer- Building applications with no-code tools like Glide and low-code platforms like Appsmith and Webflow.
- Creating API gateways and third-party connectors.
- Developing complex automations on Make.com and N8N.
- Developing complete mobile applications as PWAs (NodeJS, React, Tailwind...).
- Decision of tools to use depending on the client's needs.
- Typescript
- NodeJS
- React
- SQL
- REST APIs
- Serverless
- PWA
- Database Design
- API
- N8N
- Make.com
- Tailwind
- GlideApps
- ERP
- No-Code
- Low-Code
III Financements
- Creation of a search engine for information on French startups (structured outputs via LLM, external APIs, and scraping).
- Creation of a Chrome extension and bidirectional API bridges.
- Automation of data enrichment: LLMs, open APIs, scraping.
- LLM
- API
- Scraping
- Chrome Extension
- Backend Development
- NodeJS
- Typescript
EPITA
- Creation of programming practicals for first-year EPITA students.
- In-person teaching of C# and OCaml languages.
- Writing subjects in English.
- C#
- OCaml
- Teaching
- English
French Chess Federation
Education
EPITA
Current Employer- Major in SCIA (Cognitive Sciences and Artificial Intelligence) with NLP (Natural Language Processing) specialization.
- Program: Convex Optimization, Data Engineering, Machine Learning, Deep Learning, Advanced NLP, MLOps, Big Data, AI and Cognitive Sciences, Automatic Speech Recognition.
- Minor in Robotics & AI: Robotic architecture with ROS2 and reinforcement/imitation learning.
- Integrated preparatory classes followed by engineering curriculum specialized in Artificial Intelligence with a minor in Natural Language Processing.
- International English section (courses entirely in English).
- Direct admission, "Grand classé" in the Advance competition.
- AI
- NLP
- Machine Learning
- Deep Learning
- MLOps
- Big Data
- Data Engineering
- Convex Optimization
- Automatic Speech Recognition
- Robotics
- ROS2
- Reinforcement Learning
- Python
- Typescript
- React
- Java
- C
- OCaml
University of Seoul
- Semester abroad as part of my studies at EPITA.
- Courses taken: Machine Learning & Labs, Computer Architecture, Discrete Mathematics, Special Topics in Korean (Study of Korean culture), Beginning Korean Language.
- Machine Learning
- Computer Architecture
- Discrete Mathematics
- Korean
Lycée Saint-Dominique
Stack
Projects(6)
🏆 Participated in the Iterate Paris Hackathon (Source AI for Retail Track)
Built Albert, an automated pharmacy procurement and reporting system that transforms ERP exports into AI-narrated presentations:
- 📊 Automated data extraction from CSV exports with Python-based KPI computation
- 🤖 AI-powered presentation generation using Claude API for intelligent summaries
- 🎙️ Voice-synthesized narration with ElevenLabs for professional audio
- 🤝 Deployed Recall bot for live Google Meet presentations with Q&A capabilities
- 🏗️ Full-stack implementation: Python data pipeline, Node.js presentation engine, Express API, and React frontend
Bonus points earned: Claude AI integration (+2) and ElevenLabs voice synthesis (+2).
- Claude API
- ElevenLabs
- OpenAI Realtime API
- Node.js
- Express
- Python
- React
- AI Narration
- Data Processing
- Google Meet Bot
🏆 Participated in the European Defense Tech Paris hackathon
Over 48 hours, our team tackled the challenge of breaking the 1:1 operator-to-drone dependency:
- 🗺️ Built a Proof-of-Concept for a mission planning platform
- 🤖 Developed an AI Geo Agent that significantly speeds up mission preparation
- 🚁 Created a drone-agnostic system, validated in a simulated environment using the Parrot ecosystem
The hackathon focused on real military needs in the Ukraine context, with engaging talks from defense tech partners.
Created with teammates Dmytro Lubenets, Sofyen MARZOUGUI, and Quentin MARQUET.
- Mission Planning Platform
- AI Geo Agent
- Drone Systems
- Parrot SDK
- Simulation
- Defense Tech
- Multi-agent Systems
Built a benchmark to measure multimodal LLM visual reasoning inspired by TimeGuessr as a summer side project.
- Evaluated 14 frontier models (Google, OpenAI, Anthropic, Mistral, Qwen, etc.) on 30 historical images each.
- Automated scoring with the official TimeGuessr system; Gemini 2.5 Pro led with 9686/10000, closely followed by GPT-5.
- Published a detailed analysis and interactive demo so others can upload their own images.
🔗 Analysis and demo website: https://timeguessr-llm.titouan.verhille.me/.
🔗 Repository: https://github.com/titouv/timeguessr-llm,
- Multimodal LLM Evaluation
- Computer Vision
- Benchmark Design
- Data Analysis
- Scientific Communication
🏆 Finalist (Top 6 out of 37 teams) in the Mistral MCP Hackathon
Built in 24 hours, our Job Application Companion MCP integrated with Mistral LeChat:
- Intelligent job search engine to discover perfect opportunities
- Automated generation of customized resumes and cover letters using LaTeX
- Streamlined application process with automatic form filling powered by Browserbase and LLM control
Created with teammates Liza Fretel, Yuntian Shen and Weiqi ZHANG.
Organized by Cerebral Valley and sponsored by Mistral AI, Balderton Capital, Alpic, Motier Ventures, Bria AI, Qdrant, and Weights & Biases.
- Model Context Protocol
- Mistral LeChat
- LLM
- LaTeX
- Browserbase
- Job Search
- Form Automation
Social Links
LinkedIn
titouan-verhille
GitHub
titouv