Finetuning similarity search

Domain-specific embeddings

Large-scale, pre-trained language models have changed how we view natural language in computer science. Finetune transformer models to build better embeddings for your similarity search.

AI performance = 
model architecture + raw training data

With refinery, you can transform your raw data into training data in hours
- and continuously improve the quality of your data programmatically.
Shorter development cycles
Drastically reduce the time needed to transform raw data into AI training data. This allows you to iterate faster on your product development.
Auditable data
Don’t want your AI to be a blackbox? Start with documenting the data! We do so automatically through enriching your data with metadata.
High-precision models
AI performance scales with the amount of high quality training data. Build your models on a programmable data stack with refinery.

Build your AI with ease

refinery is an open-source developer tool, allowing you full customization for your task at hand.
Document Classification
Tag your documents with valuable metadata.
Learn more
Intent Classification
Automatically route questions from your clients.
Learn more
Urgency Detection
Find which messages are most urgent and quickly respond to them.
Learn more
Service Desk Automation
Streamline tasks and help clients more efficiently.
Learn more

Become a data pioneer now

We are building tools for the age of data-centric AI.
Let's build great use cases together.