Refuel launches platform to automate data labeling using LLMs, having already delivered 10B+ annotations across customers

Refuel makes clean, labeled data available to enterprises at better-than-human accuracy, 100x faster, and at infinite scale

SAN FRANCISCO, Oct. 17, 2023 /PRNewswire/ -- Refuel.ai today announced the launch of its platform, Refuel Cloud, and its large language model (LLM) purpose-built for data labeling. Refuel Cloud automates data cleaning, labeling and enrichment at scale for use cases across industries by leveraging state-of-the-art LLMs, including Refuel's own LLM. Clean data is the bedrock of all modern AI and data-driven software products, and producing this data has long been bottlenecked by human labor. Refuel Cloud makes it easy for enterprises to create large-scale, accurate datasets they need in minutes, in what would otherwise take weeks.

Refuel Cloud revolutionizes the data labeling process – teams can simply describe in natural language how they want their data labeled, and not have to do the tedious work of labeling themselves. State-of-the-art LLMs label the data in minutes, and domain experts can work within an easy-to-use interface and provide feedback when necessary. Instead of weeks of manual work, teams can create highly accurate datasets in hours and easily build applications for ongoing labeling.

This is coupled with the announcement of Refuel's own LLM, the world's first language model built for data labeling and enrichment. Refuel LLM displays impressive performance on a set of comprehensive benchmarks released today – better than human labeling accuracy and outperforming ChatGPT. It's also 8x faster and less than 10% the cost of existing state-of-the-art LLMs for the same labeling task. Refuel LLM is available as part of Refuel Cloud and can run within a customer's environment providing enhanced data security.

"Refuel empowers companies to overcome their data bottlenecks, enabling projects to get to production which would otherwise never get off the ground," said Rishabh Bhargava, Refuel.ai co-founder and CEO. "No more waiting weeks for labeled data to become available, or hiring large offshore teams for data enrichment – Refuel Cloud is orders of magnitude faster, extremely accurate and cost-effective. We started Refuel with the mission of enabling clean, labeled data at the speed of thought, and are already delivering billions of labeled data points for customers."

Refuel is used by enterprises and startups in a range of industries, including financial services, insurance, e-commerce and education. Refuel has demonstrated comparable or better accuracy to skilled human workers on tasks as varied as data extraction from insurance policy documents to toxicity classification in content moderation and categorization of financial transactions categorization. This is made possible by the flexibility of Refuel's platform to adapt LLMs to any use case across verticals.

"Data labeling is a critical need for Enigma. We compared Refuel against our existing manual data labeling solution, and not only did Refuel's LLM produce more accurate labels as validated by human teams, but their speed for large datasets was astounding – 1 day compared to 5 weeks," said Charles Zhu, VP of Product Management at Enigma. "We're excited to partner with Refuel on many more projects and leverage LLMs for data labeling and enrichment."

Refuel Cloud comes hot on the heels of the company's launch with $5M in funding and announcement of their open-source library, Autolabel. Autolabel is a product built for technical users to accelerate data labeling using LLMs. Early Autolabel users saw a 100x speedup for dataset creation and labeling at par or better than human accuracy. Now with Refuel Cloud and LLM, the company is opening their LLM capabilities to scale data labeling across enterprise use cases.

"We're thrilled to release the world's best data annotator, Refuel LLM, to automate data labeling and enrichment for every business and every use case," said Nihit Desai, Refuel.ai co-founder and CTO. "It can label your data better, faster and cheaper than human annotators. Furthermore, it can be finetuned and deployed anywhere. With data no longer a bottleneck, there is no limit to what teams can accomplish."

Refuel's founding team has previously worked at DeepMind, Meta, Apple, Uber and Amazon. They are backed by General Catalyst and XYZ Ventures, and a host of successful operators and founders from Slack, OpenAI, Upstart, Datadog and others.

To sign up for Refuel Cloud, visit: https://www.refuel.ai/get-started and access the Refuel LLM playground at https://app.refuel.ai/playground.

About Refuel AI
Refuel.ai is accelerating the era of AI abundance through clean, labeled data for enterprises. Their technology leverages state-of-the-art large language models and provides companies across a variety of verticals with a platform to tackle their unique data labeling, cleaning and enrichment needs. Headquartered in San Francisco, CA, the company recently launched out of stealth by founders Rishabh Bhargava and Nihit Desai. Investors include General Catalyst and XYZ Venture Capital. For more information visit www.refuel.ai.

SOURCE Refuel.ai, Inc.