Cleanlabs Gets $25m, Unveils LLM Output Cleaner

San Francisco-based automated data curation specialist Cleanlab has secured $25 million in Series A funding to expand its operations; and has launched new features to deal with 'unreliable outputs' from Large Language Models (LLM).

Founded in 2021 by three PhDs in Computer Science from MIT, Cleanlab promises to turn unreliable data into reliable models and insights by automatically finding and fixing errors in both structured and unstructured datasets - including visual, text and tabular data. Its software automatically analyzes and corrects outliers, ambiguous data and mislabeled data.

Flagship platform Cleanlab Studio now includes Trustworthy Language Model (TLM), available now in beta version, which it says produces high-quality LLM outputs like ChatGPT, Falcon and similar LLMs but also adds a trustworthiness reliability score to all outputs. This will help users automate reliability and quality assurance for systems that rely on LLM outputs, synthetic data and generated content.

The round, which massively extends previous total funding of $5m, was co-led by Menlo Ventures and TQ Ventures - from which companies Matt Murphy and Schuster Tanger respectively will join the Cleanlab board. Existing investor Bain Capital Ventures (BCV) and new investor Databricks Ventures also participated in the funding.

Cleanlab co-founder and CEO Curtis Northcutt comments: 'It's the culmination of over a decade of work to introduce Cleanlab Studio, which reimagines what AI and analytics can do for people and enterprises now that we can automate data curation and reliability'. Menlo's Murphy adds: 'While most of the investment in generative AI is chasing the biggest, baddest, and best model, the reality is that there is a massive complimentary opportunity that can shave billions off those efforts and lead to a better outcome. That is Cleanlab'.

The company is on the web at www.cleanlab.ai .