NLP Research Scientist, Deezer
One Year Fixed Term Contract
Office located in Rue de Calais, Paris
Audio streaming platform
Be an early applicant
Audio streaming platform
1001+ employees
Be an early applicant
Office located in Rue de Calais, Paris
1001+ employees
Company mission
Connecting fans with artists and creating ways for people to live the music.
Role
Who you are
- The candidate should hold a PhD (or equivalent research experience) and publications in NLP or in a related field
- Deep understanding of state-of-the-art methods in NLP
- Experience with training and fine-tuning language models and machine learning systems. Experience with ML on multimodal data would be a plus
- Fluent in Python, NLP libraries (transformers, NLTK, spacy…), and a deep learning framework (PyTorch, …)
- Ability to produce clean and reusable code
- Fluency in English. Knowledge of French would be a plus
- Knowledge of the Docker environment would be a plus
- Knowledge of Music Theory would be a plus
- If this role speaks to you and aligns with your interests and capabilities, even if you don’t meet 100% of the qualifications listed, please give it a try and tell us more! #BeYou
What the job involves
- Deezer is seeking an NLP research scientist with a strong academic background to join its Research team for a 1-year contract
- The team conducts long-term, exploratory research aimed at advancing knowledge representation of music content and understanding listener interactions with music
- With a catalog of several hundred million tracks and more than ten million users, enabling easy browsing, song retrieval, and music recommendations is a challenge
- Your mission will be to develop robust, language-focused, musical item (i.e., songs, artists) representations, with the goal of enabling item or user description generation (e.g., playlist captioning, user preference summary) and supporting user interaction via natural language queries. This will have direct applications in both the recommender system and the search engine of Deezer
- You will collaborate with other researchers, production teams responsible for music recommendation and search, and music curators, and have access to large-scale data on music content and usage, allowing you to have a real-world impact on millions of music listeners
- Define domain-specific evaluation schemes and collect data
- Propose, develop, and evaluate NLP systems to create a music item representation space from multimodal data
- Collaborate with other researchers and with the recommendation and search; team, sharing ideas, knowledge, and projects
- Scientific watch
- Scientific communication (papers, publications, blogposts…)
Our take
Deezer is a global music-streaming platform built to help people connect with the music they love. Listeners can explore millions of tracks, discover new artists, and enjoy features like Flow, curated playlists, and interactive tools such as Music Quiz and Shaker.
The company operates in a crowded market dominated by giants like SPotify and Apple Music, but it's leaning into personalisation to stand out. Its latest update lets users customise themes, colors, icons and layouts, and gives instant access to their favourite artists, playlists and podcasts through the new Quick Access feature. So, with this approach, it wants to make the app feel more personal in the hopes of appealing to listeners looking for something a bit more tailored.
AI-generated music is another major factor shaping Deezer's direction (and it's a complicated one). Around 50,000 fully AI tracks are uploaded daily, and while many users may be curious, most want them clearly labelled and worry about their impact on real artists. Deezer already detects and labels AI content and keeps it out of recommendations and fraudulent payouts. If it can continue to handle this transparently, AI could become a trust-building advantage rather than a threat for the company.
Freddie
Company Specialist at Welcome to the Jungle
Insights
11% employee growth in 12 months
Company
Funding (last 2 of 7 rounds)
Aug 2018
$186.9m
LATE VC
Jan 2016
$116.8m
SERIES E
Total funding: $551.9m
Company HQ
Saint-Georges, Paris, France
Leadership
Alexis Lanternier
(CEO)Previously co-founded Essor where they served as the President for nearly 4 years. Prior to that, they were an EVP of eCommerce at Walmart Canada.
Salary benchmarks
We don't have enough data yet to provide salary benchmarks for this role.
Submit your salary to help other candidates with crowdsourced salary estimates.
Share this job
View 17 more jobs at Deezer