Jean-Michel D.

✨ You’ll find:
- Why the bottlenecks aren’t about AI/ML
- How to think about ROI (why, how much, how long?)
- Practical tips to make projects more robust

👆Link in the thread Curious: what’s the biggest blocker you’ve seen in AI projects?

www.the-odd-dataguy.com/2025/09/06/d...

#AI #MachineLearning #DataScience #ROI #ProjectManagement #TechLeadership #AIStrategy

🚨 95% of AI projects fail!? Not really. The truth is more nuanced—and more interesting.

MIT’s NANDA report made headlines, but the reality of running AI projects is more complex (and many points are missing).

I wrote a piece to share my take, plus lessons from deploying and operating these systems

www.the-odd-dataguy.com/2025/06/01/k...

two men are sitting at a table and one of them is saying ah non les noix ALT: two men are sitting at a table and one of them is saying ah non les noix

Salut !
Je plonge en ce moment dans le deep learning, le NLP et l’IA générative — mais marre des datasets pas fun.
J’en ai créé un à partir des scripts de Kaamelott ⚔️
Parsing & prompting avec #DeepSeek, #OpenAI et #Mistral.

📎 Article + dataset 👇 Vos retours ?
#Kaamelott #LLM #NLP #IA

www.the-odd-dataguy.com/2025/06/01/k...

two men sitting next to each other with the words ca c'est bien ca on the bottom ALT: two men sitting next to each other with the words ca c'est bien ca on the bottom

Hi, I've been diving into deep learning, NLP, and generative AI recently, but I was tired of not having fun datasets. I built one from Kaamelott scripts and used #DeepSeek, #OpenAI, and #Mistral LLMs for parsing and prompting.

📎 Article + dataset 👇 — feedback welcome!
#Kaamelott #LLM #NLP #AI

PS: Writing this definitely gave me a few ideas for new deep dives, but I’d love to hear your thoughts! What stood out to you? Is there anything you'd like me to explore further? 👇

My Decade's Journey in Data and AI The Odd Data Cast · Episode

🎧 Audio version: open.spotify.com/episode/1fi0...

📖 Article: www.the-odd-dataguy.com/2025/02/13/1...

#DataScience #MachineLearning #AI #ArtificialIntelligence #BigData #DataAnalytics #MLOps #TechLeadership #CareerGrowth #CareerDevelopment

🖊️ Don't expect a super technical deep dive or industry gossip. Instead, it’s my personal experience, thoughts, and references from a decade in the field. Hope you enjoy it!

📔 It’s been an on-and-off project, but over the past few weeks, I finally found the time to wrap up my notes from the past months. It’s dense—probably my longest article—so buckle up!

🎢 The past year has been a rollercoaster, with a lot happening around me. Somewhere along the way, I realized that 2024 marks 10 years of working in data & AI. Back in June, I thought it would be a great idea to share some takeaways from this journey.

Socially-Motivated Music Recommendation - Spotify Research tldr - The wide variety of reasons people listen to music include both individual motivations, like self-awareness and mood regulation, as well as social motivations, such as demonstrating belonging t...

The Spotify pepites:

- Socially-Motivated Music Recommendation (research.atspotify.com/2024/06/soci...)

- Exploring Local Music’s Place in Global Streaming (research.atspotify.com/2024/05/expl...)

ACM recsys 2024 : My recap with my selection (www.the-odd-dataguy.com/2024/11/25/r...) and my top pick is A Pre-trained Sequential Recommendation Framework: Popularity Dynamics for Zero-shot Transfer (arxiv.org/abs/2401.01497)

Track Mix Generation on Music Streaming Services using Transformers This paper introduces Track Mix, a personalized playlist generation system released in 2022 on the music streaming service Deezer. Track Mix automatically generates "mix" playlists inspired by initial...

- Track Mix Generation on Music Streaming Services using Transformers(arxiv.org/abs/2307.03045)

- Personalised Recommendations for the BBC iPlayer: Initial approach and current challenges (dl.acm.org/doi/10.1145/...)

For the @bsky.app crowd not on linkedin , there is maybe a few interesting extras for this super list

ACM recsys 2023 : My recap with my selection (www.the-odd-dataguy.com/2023/11/13/r...) and my top picks are:

🧮 In this article, I explain how matrix factorization relates to accurately summarizing user feedback, provide a historical overview of various factorization techniques, and share some considerations and tips for effectively using matrix factorization in real projects.

🖊️ Last summer, while working on a few projects, I revisited the classic matrix factorization technique to create 2D representations of users. It turned out to be such an interesting topic that I decided to write an article about it—so here we are!

www.the-odd-dataguy.com/2024/08/11/r...

#HolidaySeason #HappyHolidays #HiPPOEffect #ABTesting #DataDriven #GrowthMindset #KinderHappyHippo #StockingStuffers #DataScience #DecisionMaking #ProductTesting #LeadershipTips #WorkLifeBalance #SeasonGreetings #Teamwork #CelebrationTime

Wishing everyone a joyful holiday season filled with celebration, family, friends, and self-care. 🎄✨

P.S. If you’re wondering: do Happy Hippos taste like Kinder Bueno? Spoiler alert—they don’t! 😊

🧦Last week, while grocery shopping for stocking stuffers, I couldn’t resist grabbing this Happy Hippo from Kinder. It made me smile, thinking of the good old HiPPO effect (highest-paid person's opinion - (growthmethod.com/hippo-effect/) — a familiar concept for anyone in A/B testing

This article outlines my journey in building this playground. I'll discuss setting up the gym and developing the initial agents, which are crucial steps in teaching an AI to navigate and succeed in this environment.

#MachineLearning #GameDevelopment #AI #Tech #Innovation #Gaming

📦 First things first: I needed to create a sandbox environment where an agent could learn and eventually master the game. I wrap my first experiment in an article in June

www.the-odd-dataguy.com/2024/06/23/s...

🍉Last winter, I stumbled upon a game called Suika, a fun mashup of Tetris and Puzzle Bobble. It's a lot of fun, and I thought it would be the perfect opportunity to explore the world of game agents.

Features and Design Principles of a Recommender System The Odd Data Cast · Episode

🎧 If you prefer an audio version of the article, there's an AI conversation summary built on top of it using NotebookLM from Google.

Enjoy the reading/listening!

open.spotify.com/episode/0DTi...

🧐This work didn’t come out of a vacuum. I built it on top of insights from Amazon Web Services (AWS), Recombee, Spotify, Netflix, and more, complemented with a dash of my own experience/salt🧂

#RecommenderSystems #MachineLearning #DataScience #Personalization

📝Last April, I was preparing a workshop and boot camp on recommender systems with some of my colleagues at @ubisoftgames.bsky.social, an article emerged, focusing on the essential features of these systems and key considerations for their design.

www.the-odd-dataguy.com/2024/04/07/f...

a blue background with a white butterfly in the middle ALT: a blue background with a white butterfly in the middle

Hello Bluesky! 👋

It looks like I’ve finally landed where many of my favorite data/ML folks migrated after the whole X mutation. Excited to switch up my watch and explore new horizons under this blue sky! 🌤️🟦

I’ll also be (re)sharing some of my recent thoughts and takes on the topics - stay tuned!