Regular updates on the latest VC-backed AI startups. Follow along to stay informed!
LlamaIndex, a data framework for large language models, raised $8.5M in Seed Round funding. The round was led by Greylock, with angel investors including Jack Altman, Lenny Rachitsky, Mathilde Collin (CEO of Front), Raquel Urtasun (CEO of Waabi), and Joey Gonzalez (Berkeley) participating.
Problem to be Solved
“Many users building applications on top of LLMs want to unlock new use cases with their own private data,” is how LlamaIndex CEO Jerry Liu puts it. This is a familiar story at AI Deal Watch - most model orchestration tools (like LangChain, which we wrote about here in April) aim to connect LLMs to data outside their training set.
How They Use AI
LlamaIndex is a tool that enables developers to connect AI language models to databases. It is very similar to LangChain but has a stronger focus on data integration and less of a focus on user-facing solutions and prompt management. Competition for this use case is stiff from LangChain and vector database companies like Pinecone.
Business Model
TBD, but LlamaIndex will likely follow the open-source dev tools model of monetizing highly integrated deployment and hosting solutions popularized by HuggingFace and others. The company will use the funds to build an enterprise offering on top of its free, open-source project, which has 15K Github Stars, 19K Twitter followers, 200K Monthly downloads, and 6K Discord users. Pre-revenue companies raising on these vanity metrics of developer popularity is starting to feel like bubble behavior (crypto flashbacks anyone?), but teams at companies like Uber, Instabase, and Front are using LlamaIndex today, providing evidence of a real use case.
Granica, an AI efficiency platform, raised $45M in Seed Round funding. NEA led the round, with participation from Bain Capital Ventures and several individual investors, including former Tesla Inc. CFO Deepak Ahuja, Eventbrite Co-Founder Kevin Hartz, and Okta Co-Founder Frederic Kerrest.
Problem to be Solved
Developing AI foundation models requires vast training data, which is expensive to store and use. Granica claims to improve cost efficiency by enabling AI teams using AWS S3 and Google Cloud Storage to derive maximum value from their growing volumes of training data.
How They Use AI
Granica claims to reduce the size and cost of petabyte-scale AI training data in cloud object stores by up to 80% using novel compression and deduplication algorithms. In the context of data reduction with Granica Crunch, users’ applications write data into buckets as usual (no changes) and that data is automatically and losslessly reduced. The applications then read data, whether reduced or not, via the Granica S3-compatible API/SDK. Applications can optionally write data inline through the Granica API/SDK to further increase efficiency and savings.
Business Model
Pricing for the service is unique, with Granica offering an outcome-based pricing model that only charges users a small percentage of the savings generated per month (from a reduction in storage needs, lower data transfer costs, or increased efficiencies in AI training processes) so that deployment of the solution delivers upside only. This is a rare example of business model innovation in AI, as the recent wave of startups has been driven almost exclusively by technology innovation.
EvenUp, an AI-driven support platform for personal injury lawyers, raised $50.5M in Series B funding. Bessemer Venture Partners led the round, with participation from Bain Capital Ventures, Behance Founder Scott Belsky, and legal tech leader Sameer Dholakia.
Problem to be Solved
EvenUp claims that over 90% of the millions of injury cases filed every year are settled privately, leaving plaintiffs and injury attorneys uncertain about the true value of their specific cases. The company aims to ensure plaintiffs receive rightful compensation with an AI product that analyzes medical records and generates demand letters.
How They Use AI
We believe EvenUp is training in-house NLP models for information retrieval and entity recognition. It’s likely that the company uses an ensemble of in-house models to extract relevant data for the demand letter and uses an external model (e.g., GPT-4) to summarize the findings into natural language.
Business Model
SaaS. EvenUp sells software subscription licenses to personal injury lawyers, and the platform is used by over 300 firms today. This is a notoriously difficult target customer (especially considering EvenUp’s GTM is not product-led) because most law firms are small, which results in low average contract values.