# Signeasy AI We're the AI team at [Signeasy](https://signeasy.com) — building the machine-learning systems that turn unstructured contract data into something a sales rep, a paralegal, or a finance lead can search, summarise, and act on in seconds. ## What we work on - **Document AI** — extracts contract types, key terms, and signing metadata from PDFs at ingest time, so the rest of the product can treat a contract as structured data instead of a binary blob. - **Natural-language search** — turns one English query (*"top 10 NDAs by value"*, *"expiring contracts over $500K"*) into a structured filter spec the contract-management UI dispatches into its existing data layer. Resolution-only — no LLM in the hot path. - **Contract categorisation** — classifies a freshly-uploaded document into one of ~20 first-class contract types so the right extraction pipelines fire automatically. ## Models hosted here | Model | Purpose | |---|---| | [`Signeasy/nl-search-joint-bert`](https://huggingface.co/Signeasy/nl-search-joint-bert) | DistilBERT-based JointBERT classifier (intent + slot + span heads) powering natural-language search over contract data. Private. See the model card for eval results, OOD defences, and version history. | ## How we work Most of our models are deliberately small — DistilBERT, MiniLM, distilled cross-encoders — so they run on CPU at the latency the product needs. Search has to feel instant; an end-user doesn't wait. We use Hugging Face Hub as our model registry: training happens in our research repo, the artifact gets pushed here as a private repo at a tagged revision (`v3.5`, `v3.6`, …), and production pulls it via `snapshot_download` at image-build time so cold starts are self-contained. ## Audience This org is primarily for the Signeasy AI team and the engineers integrating our models into the product. The repos here are not a general-purpose release; they are the internal artifact registry. If you're a Signeasy engineer who needs access to a private repo here, ping the AI team on Slack.