Research
Building State-of-the-Art Models
We train models , publish our findings, and push the boundaries of what's possible with embeddings and LLMs.
Join Research
35+
ML Engineers
6
Papers Published
8
Models in Production
2
GPUs Training
Research Areas
From embeddings to long-context LLMs, we're working on the hardest problems in AI.
Multilingual Embeddings
ActiveTraining embedding models that work across 100+ languages.
Legal Document Processing
ActiveEfficient architectures for processing extensive legal documents and contracts.
Inference Optimization
ActiveQuantization, speculative decoding, and custom CUDA kernels for sub-20ms latency.
Domain Adaptation
ActiveTechniques for adapting foundation models to specialized domains like legal and finance.
Check out our latest papers on arXiv and our technical blog.
Follow UpdatesJoin the Research Team
We're hiring ML engineers and researchers. Work on hard problems. Ship real products.