Florent Daudens

Content | Innovation | Management

Passionate about AI and technology trends, driven by news and enriched by human connection, I've been shaping the digital evolution of major media outlets for 15+ years.


Experience

As the current Director of Newsgathering and Deployment for national and international news at Radio-Canada, I orchestrate operations across five international bureaus, parliamentary bureaus in Quebec and Ottawa, and news journalists in Montreal.

In my former position as News Director at Le Devoir, I led a sweeping digital transformation. This involved overseeing editorial teams, initiating specialized units in video and data visualization, launching a variety of innovative products such as applications, newsletters, and election coverage, venturing into new coverage domains, and driving artificial intelligence initiatives.

Simultaneously, I have the opportunity to instruct digital journalism at the University of Montreal's Graduate Diploma in Journalism program. I remain committed to fostering growth in our profession, with contributions to the board of FPJQ and Trente magazine, as well as involvement in groups like Hacks/Hackers Montreal. My diverse experience spans roles at La Presse, Rue89, France3, and the Federation of Quebec Journalists.

Click here to view my detailed resume.


Projects

💬📚 An AI-Assisted Course

In the "Digital Journalism" course at UdM, this year I am integrating a conversational robot as a course assistant. Using the Retrieval-augmented generation (RAG) method, this robot is designed to supplement the teachings with relevant information extracted from course transcripts.

This educational project allows students to interact directly with an advanced language model, thereby enriching their learning experience. I collect their feedback and questions in the system, exploring the impact of this AI assistant on the evolution of the course. A shared learning journey for both me and the students.

On the technological side, this program is based on GPT-4, LangChain, LangSmith, and Streamlit.

To try out the tool, click here. For more details on this project, read this post.

The AI Gazette

With a curiosity to explore ChatGPT's programming abilities, I crafted a program using Python, FeedParser, and OpenAI's API.

Hosted on AWS for automation, this program sends out an email at 8 a.m. daily, providing a summary of the past 24 hours' AI-related articles using text-davinci-003. You can find more details in this post.

🤖🗳️ Resultbot, the Election Results Robot

Three objectives for this project:

  • Test the capabilities of large language models to interact with numerical data
  • Evaluate the agents' reasoning abilities to answer a question
  • Record user interactions to improve prompts and potentially fine-tune a model

For those who like to look under the hood, I've tested several solutions. The final configuration relies on a Pandas agent from LangChain, GPT-4 for the language model, Langsmith for interaction tracking, and Streamlit for deployment.

To try out the tool, click here. For more details on this project, see this post.

📰🧵 AI-Assisted Twitter Thread Generator

A simple tool to flood the Internet generate a Twitter thread (or the 24th letter of the alphabet) from news articles, and an opportunity to test LangChain Hub for prompts. This project uses:

  • Streamlit for the user interface
  • OpenAI GPT-3.5 Turbo for text generation
  • LangChain Hub for access to predefined models and prompts
  • Newspaper3k for online article extraction

To try out the tool, click here. For more details on this project, read this post.


Certificates

Artificial Intelligence
  • Large Language Models with Semantic Search - DeepLearning.AI
  • LangChain for LLM Application Development - DeepLearning.AI
  • ChatGPT Prompt Engineering for Developers - DeepLearning.AI
  • AI For Everyone - Coursera

Data Visualization
  • Data Visualization and Infographics with D3 - Knight Center
  • Doing Journalism with Data - European Journalism Center
  • QGIS - Knight Center