Document search
An index, embeddings, answers grounded in your files, with the source named.
Artificial intelligence
Artificial intelligence inside the product.
Search across your documents, an assistant for your trade, transcription, extraction. The model is chosen for the task, hosted according to the confidentiality constraint, and connected to your data.
Delivered uses.
Assistants
An agent that knows your tools and stops when an action needs approval.
Documents and voice
Reading papers, recognition, speech synthesis, when the work is spoken or on paper.
Hosting the models
Inference on your side or ours, queues, cost and latency kept visible.
Techniques.
Python for models and data, Go or Rust for the services around them, PostgreSQL and vector indexes for search. Open models or APIs, depending on what the data is allowed to leave.