Artificial intelligence startup Fireworks AI, a platform that helps companies fine-tune and deploy generative AI models for specific purposes, today announced it has raised $52 million in a Series B funding round.

The round was led by Sequoia Capital which placed its valuation at $522 million. Notable investors joining the round included Nvidia Corp., Advanced Micro Devices Inc. and MongoDB Inc. alongside previous investors such as Benchmark and Databricks Ventures. Former Snowflake Chief Executive Slootman, former Meta Chief Operating Officer Sheryl Sandberg, Airtable CEO Howie Liu and Scale AI CEO Alexandr Wang also joined in the fundraise. The new funding brought the total raised by the company to $77 million.

Fireworks provides developers with access to what it says is the fastest and most cost-effective inference for popular models that they can fine-tune and optimize for their own purposes. Among the 100 state-of-the-art models for text, image, audio, embedding and multimodal formats, the company delivers specialized models, including Llama 3, Mixtral and Stable Diffusion.

“Unlike proprietary mega models that are generic, non-private, and hard to customize, Fireworks AI provides smaller, production-grade models that can be deployed privately and securely,” said CEO Lin Qiao. “Using minimal human-curated data, our ultra-fast LoRA fine-tuning allows developers to quickly customize models for specific needs, transitioning from dataset preparation to querying a fine-tuned model in minutes. These fine-tuned models are seamlessly deployed, maintaining the same performance and cost benefits as our base models.”

Since launching two years ago, Fireworks has made major upgrades to its platform and inked numerous deals with providers in the AI stack. These include partnerships with Nvidia, AMD, Amazon Web Services, Google Cloud Platform and Oracle Cloud for model infrastructure optimization. The company also partnered with the cross-platform NoSQL database MongoDB to create an interactive retrieval-augmented generation or RAG product, which helps bring authoritative real-time data into large language models to increase accuracy and reduce errors.

The company said it intends to double down on researching and building compound AI systems, which use multiple components instead of a single large mega-sized model. Compound AI systems take on tasks using a myriad of interacting parts, including multiple smaller models, external tools modalities and elements that interact similarly to microservices. Collectively, the company says, they get a job done.

Approaching this vision, Fireworks recently introduced FireFunction V2, a function-calling AI model that can orchestrate multiple other LLMs and their capabilities, external data sources, search, transcription and other application programming interfaces. It can do all this while also acting as a chatbot or preserving other user-facing AI capabilities.

“While leaderboards emphasize larger models, real-world AI results, especially in production, increasingly come from compound systems with multiple components,” said Qiao.

Superhuman Labs Inc., the provider of AI-powered email services, used Fireworks to create its Ask AI. The system is a compound AI system designed to deliver answers from a user’s email inbox with conversational queries and interactions.

Benny Chen, co-founder of Fireworks, and Gregory Maxson, senior director and global lead of AI GTM and strategic partners at MongoDB, spoke with theCUBE’s Research Chief Analyst Dave Vellante in May during a broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how Fireworks enables innovation and the partnership with MongoDB.

Image: Pixabay