We’re a small team of researchers and operators who care deeply about understanding how AI systems work, how they fail, and where careful human judgment can make them better.
About GroundTruth
Who we are
What we do
We test AI systems for labs and enterprises, with ready-to-license benchmarks and custom evaluations for specific models, domains, and use cases.
We also create, enrich, and quality-check training data across multiple modalities, for every stage of your AI training pipeline.
We work with a network of vetted domain experts and generalist annotators to provide the human expertise behind this evaluation and training data.
Team
Miroojin BakshiCo-founder & Research Head
Mohit KhullarCo-founder
Aman GoutamFounder
Connect
If you're a lab or enterprise looking to license a benchmark, commission a custom evaluation, or get help with training data, or if you just want to talk, book a call or email
Domain experts and generalist annotators looking to get paid for applying human judgment in our projects can join our Expert Platform.
Our benchmarks and data are on Hugging Face. Follow us on LinkedIn and X for company updates.