Launched in February 2023, Arcee.ai specializes in developing and deploying domain-adapted language models (LLMs) and small language models (SLMs) tailored to specific enterprise domains. The Arcee platform enables the merging and deployment of the domain-adapted language model system (DALM), which involves a three-step process: Deploying pre-trained, in-domain models optimized for specific verticals; training a retriever model on contextual data; and fine-tuning the LLM through an instruction-based approach. This results in LLMs, harmoniously synchronized with a company's domain, enhances language understanding and generation capabilities.
The company has developed custom SLMs, including a finance model (Llama-3-SEC) and a general knowledge model (Arcee-spark), which can be deployed for a range of applications across industries in healthcare, legal, finance, and insurance. Arcee.ai's MergeKit-powered platform demonstrated improvements in domain adaptation, particularly in the medical and legal verticals.
The firm claims its domain-specific approach can deliver 20x cost savings on training of LLMs. In September 2024,Arcee launched SuperNova, a 70-billion parameter language model designed for enterprise deployment. Earlier in January 2024, the company developed a US patented, end-to-end, domain-trained model, showcasing a 50% improvement in retrieval over baseline models.
Key customers and partnerships
As of August 2024, Arcee.ai's customer base included notable firms such as Thomson Reuters, AWS, Intel, and Guild.
The company has partnered with AWS, Hugging Face, Intel, MongoDB, and TogetherAI.
In January 2024, Arcee.ai partnered with Mergekit, whose founder, Charles Goddard, also joined the team. As of April 2024, MergeKit garnered over 3,000 GitHub stars.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.