Draft:AICrowd

• Comment: In accordance with the Wikimedia Foundation's Terms of Use, I disclose that I have been paid by my employer for my contributions to this article. Sananavati (talk) 14:14, 15 May 2025 (UTC)

---

AICrowd is a Swiss-based platform for crowdsourcing artificial intelligence (AI) solutions through open challenges. Founded in 2017 as a research initiative at the École Polytechnique Fédérale de Lausanne (EPFL), it was initially known as crowdAI before becoming AICrowd SA. AICrowd enables data science experts and enthusiasts to collaboratively solve real-world problems through challenges.

History

AICrowd was co-founded by Sharada Mohanty and Marcel Salathé. The platform emerged from EPFL to support reproducible machine learning research and broaden access to applied AI problem-solving. Early initiatives included applying AI to plant disease detection, train scheduling in collaboration with SBB, Deutsche Bahn, and SNCF through the Flatland Challenge, and the development of reinforcement learning agents in a musculoskeletal environment via the Learning to Run challenge.

Projects and Collaborations

As of May 2025, AICrowd has hosted over 320 challenges in collaboration with more than 80 academic and industry partners, distributing over US$1.4 million in cumulative prizes. Notable collaborators include OpenAI, DeepMind, Carnegie Mellon University, Microsoft, Amazon, Spotify, Stanford University, Meta, and Sony Research.

Prominent challenges hosted include:

• The Procgen Benchmark and MineRL Competitions, in partnership with OpenAI, DeepMind, and Carnegie Mellon University, part of NeurIPS 2020.
• The Million Playlist Dataset Challenge with Spotify Research, focused on music recommendation using large-scale playlist data.
• The NeurIPS 2020 Reinforcement Learning Challenge, co-organised with Amazon Science.
• Other challenges such as the Food Recognition Benchmark, Learning to Run, and the Flatland Railway Simulation have contributed to advancements in computer vision and reinforcement learning.

Community

As of May 2025, AICrowd’s community comprises over 287,000 members across more than 85 countries. Participants range from students and early-career researchers to professionals in academia and industry.

Methodology

AICrowd’s methodology emphasises openness, reproducibility, and applied problem-solving. The platform supports modular experimentation, allowing participants to submit complete machine learning pipelines with reproducible code and structured datasets. Leaderboard-based evaluation and automated testing are used to benchmark solutions, and many challenges are designed to promote collaborative learning and open research.

Values

The platform promotes responsible AI development by prioritising transparency, inclusivity, and public understanding of AI systems. AICrowd designs challenges that are accessible to contributors across skill levels, with an emphasis on educational and real-world impact.

Recognition

AICrowd-hosted challenges have contributed to over 60 research papers presented at conferences including NeurIPS, ICLR, and CVPR. AICrowd has been covered by independent media sources for its role in AI research and public-interest applications. A 2021 article by SwissInfo highlighted AICrowd as part of Geneva's ecosystem of organisations using AI to address global challenges. Geneva Solutions included AICrowd in its reporting on open research and climate-focused technologies. The platform is also listed in the EU-Startups directory of notable European AI ventures.

Corporate

AICrowd SA is headquartered in Lausanne, Switzerland. It operates independently and maintains collaborations with universities, international organisations, and technology companies.

See also

• Crowdsourcing
• Machine learning
• Reinforcement learning
• Data science competitions
• NeurIPS
• École Polytechnique Fédérale de Lausanne
• Marcel Salathé

External links

• Official website