Headquarters
San Francisco, CAFounded
2021CEO / Founder
Devvret RishiEmployees
11-50Funding
Acquired — $28.45MProducts
0 products listedWebsite
predibase.com/Profile updated
Predibase was founded in 2021 by Devvret Rishi (CEO), Piero Molino (Chief Science Officer), and Travis Addair (CTO), drawing on their backgrounds in machine learning infrastructure at companies including Uber and Google. The company is headquartered in San Francisco, California.
The platform enables teams to fine-tune open-source language models such as Llama and Mistral using LoRA (Low-Rank Adaptation) techniques, allowing multiple customized adapters to run on a shared base model simultaneously. This architecture reduces GPU costs compared to hosting separately fine-tuned models. Predibase served enterprise customers across software, fintech, and healthcare with APIs covering training, deployment, and inference.
Predibase raised $28.45 million across two rounds: a $16.25 million Series A led by Greylock in May 2022, and a $12.2 million extension led by Felicis Ventures in May 2023. The company employed approximately 25 people at the time of acquisition. In June 2025, Rubrik announced its agreement to acquire Predibase in a deal reportedly valued between $100 million and $500 million, with plans to integrate the technology into Rubrik's Annapurna AI platform for agent monitoring and governance.
Dated, source-cited figures. Estimates are marked.
No products listed yet for this company.
Who Predibase competes with, and how they differ.
Runs a public serverless inference API and hosted fine-tuning service (including LoRA) across 100+ open-source models, competing on model-catalog breadth and self-serve API access rather than Predibase's more enterprise-deployment-focused fine-tuning product.
1 product reviewed on TopReviewedOffers serverless inference plus LoRA-based fine-tuning for open-source and custom LLMs with a focus on low-latency serving speed, directly overlapping with Predibase's fine-tune-and-serve stack but without Predibase's VPC/on-prem deployment options.
1 product reviewed on TopReviewedFocuses on deploying and autoscaling custom or fine-tuned models of any type (not LLM-specific) on its own inference runtime, overlapping with Predibase's serving layer but with no built-in LoRA fine-tuning workflow.
1 product reviewed on TopReviewedLets developers run and fine-tune open-source models through a simple hosted API billed per second of compute, skewing toward prototyping and indie developers versus Predibase's enterprise fine-tuning contracts.
1 product reviewed on TopReviewedCommercializes the open-source Ray framework for distributed model training and serving (including LLMs via RayLLM), competing on self-managed, large-scale infrastructure rather than Predibase's packaged LoRA fine-tuning product.
1 product reviewed on TopReviewedGeneral-purpose serverless GPU compute for arbitrary Python workloads; teams build their own custom training/inference pipelines on Modal rather than using Predibase's packaged fine-tuning and serving product.
1 product reviewed on TopReviewedKey milestones in Predibase's history, with sources.
Rubrik agrees to acquire Predibase
Rubrik, Inc. (NYSE: RBRK) agreed to acquire Predibase, which will become a wholly owned subsidiary; disclosed in Rubrik's SEC Form 8-K. Certain Predibase shareholders representing ~77% of consideration elected Rubrik Class A stock at signing, with 77-90% of total consideration expected to be paid in stock. CNBC reported the deal price at between $100 million and $500 million (not officially disclosed by either company).
sec.govLaunches LoRA Land: 25 fine-tuned open-source LLMs
Predibase released LoRA Land, a collection of 25 open-source LLMs fine-tuned with its LoRAX framework for specific tasks (e.g. summarization, code generation), claiming they could rival or outperform GPT-4 on those tasks. Predibase said the 25 models were fine-tuned at an average GPU cost of under $8 each.
siliconangle.comRaises $12.2M Series A extension led by Felicis Ventures
Predibase announced a $12.2 million expansion to its 2022 Series A, led by Felicis Ventures, bringing total funding to $28.45 million. The round coincided with general availability of LLM fine-tuning and deployment features, including its LudwigGPT model and a "Data Science Copilot."
techcrunch.comExits stealth with $16.25M Series A led by Greylock
Predibase emerged from stealth with $16.25 million in Series A funding led by Greylock, with participation from The Factory and angel investors including Kaggle CEO Anthony Goldbloom and former Intel AI COO Remi El-Ouazzane. Co-founders Piero Molino and Travis Addair had created and maintained Uber's open-source Ludwig and Horovod ML frameworks; the founding team also included former Google Cloud AI product manager Devvret Rishi and Stanford professor Chris Ré.
techcrunch.comPredibase founded by Devvret Rishi and co-founders
Devvret Rishi co-founded Predibase in 2021, initially serving as chief product officer before taking over as CEO in 2023.
crn.comBrowse multi-perspective AI panel reviews across hundreds of AI tools, agents, and platforms. Find the right software with insights from CTO, Developer, Marketer, Finance, and User perspectives.