Predibase

About Predibase

Predibase offers a low-code artificial intelligence platform for developers and data scientists to build, fine-tune, and deploy custom AI models. The company was founded in 2020 by Piero Molino (CEO), Travis Addair (CTO), and Devvret Rishi (CPO). The platform is built upon open-source foundations, including Ludwig, a declarative deep learning framework created by Molino at Uber, and Horovod, a distributed training framework led by Addair. The initial idea, born from Molino's and Addair's experiences at Uber AI, was to productize the concepts from Ludwig to make deep learning more accessible to engineers and analysts, not just data scientists.

The founders bring a wealth of experience from major tech firms. Piero Molino was a research scientist at Uber AI, worked at IBM Watson, and founded a startup called QuestionCube. Travis Addair was a senior software engineer at Uber, specializing in distributed deep learning systems. Devvret Rishi previously served as a product manager at Google Cloud AI, including work on Kaggle and Vertex AI, and holds a Master's in Computer Science from Harvard. The founding team aimed to simplify the complex process of productionizing machine learning, which they had seen firsthand at companies like Uber, Apple, and Google.

Predibase's business model centers on providing its declarative machine learning platform as a service, targeting developers and data practitioners. It enables users to fine-tune and serve open-source Large Language Models (LLMs) and other deep learning models efficiently. The platform uses a declarative approach, meaning users specify what they want to achieve in a simple configuration file, and the system handles the complex underlying processes. This significantly reduces development time from months to days. The company generates revenue through its commercial platform, which can be deployed in a customer's private cloud or used via the secure Predibase cloud. Its clients range from Fortune 500 companies to high-growth startups like Checkr and Koble.ai.

The platform's key features include efficient fine-tuning through techniques like LoRA (Low-Rank Adaptation) and cost-effective model serving with tools like LoRAX, which allows hundreds of custom models to be served on a single GPU. This approach allows companies to create specialized models that can outperform large, general-purpose models like GPT-4 at a fraction of the cost. Predibase has raised a total of $28.4 million in funding over multiple rounds, with investors including Greylock and Felicis Ventures. In June 2025, the company was acquired by Rubrik.