Versashore Labs is our dedicated research division, working on the hardest problems in language AI โ from multilingual understanding to reliable, grounded reasoning.
The world's largest multilingual AI research program. Over 3,000 researchers in 119 countries collaborated to build the most inclusive multilingual dataset ever assembled. Aya supports 23 languages at state-of-the-art quality.
Read the paper โOur research into reducing hallucination through retrieval-augmented generation, citation tracking, and model-level uncertainty quantification. Enterprise AI must be trustworthy โ our research is making that possible.
Read the paper โA landmark study revealing how popular AI benchmarks fail to predict real-world performance. We propose new evaluation frameworks that better capture what enterprise AI actually needs to do.
Read the paper โResearch into differential privacy, federated learning, and other techniques that allow model customization without exposing sensitive training data. Enabling enterprise AI that learns from your data safely.
Read the paper โWe publish our research, release open datasets, and collaborate with the global academic community to advance AI knowledge for all.
Annual grants and mentorship for early-career researchers working on multilingual NLP, AI safety, and related areas. Applications open each spring.
Up to $50,000 in compute credits and funding for impactful AI research projects at universities and nonprofits. Apply anytime.
We partner with universities, governments, and NGOs to advance AI research that benefits everyone.