Language models working on records
Refuel.ai builds tools that use language models to clean, label, enrich and restructure data. Its documented workflow starts with uploading records or connecting cloud storage, then defining a transformation in natural language or choosing a prepared task. Users can give feedback, fine-tune a model and deploy the resulting workflow for batch or realtime processing. [1]
This places Refuel in the data layer of AI applications. The output is a transformed dataset or stream of records that another system can use, rather than only a conversational answer. The feedback and deployment steps matter because the task needs to work repeatedly over many records. The documentation establishes those functions; its claims of superior quality are the vendor’s own. [1]
Joining the inference provider
Together AI announced the acquisition on May 15, 2025. In their newsletter the following day, Refuel founders Rishabh Bhargava and Nihit Desai explicitly confirmed that the company had been acquired and that its team and technology were joining Together. This supports the ownership relationship recorded here. [3][2]
The founders described Refuel LLM-2 and its data-ingestion tools as part of that combination. Their account connects Refuel’s specialized processing of business data to Together’s infrastructure for serving and adapting models. It confirms the completed acquisition, while the longer-term integration benefits remained the companies’ stated direction. [2]
Sources
- Overview
Refuel.ai · retrieved Oct 23, 2026
- We've been acquired!
Refuel.ai · May 16, 2025
- Together AI acquires Refuel.ai to unlock data for developers and businesses building production-grade AI applications
Together AI · May 15, 2025