Abu Dhabi-based AppliedAI announces world’s first 'Large Work Model'
G42-funded AppliedAI hopes the LWM will revolutionise regulated industries
#UAE #LLMs - Abu Dhabi-based AppliedAI has launched Opus, a Generative artificial intelligence platform powered by a first-of-its-kind Large Work Model (LWM) and a proprietary Work Knowledge Graph (WKG). Designed for regulated sectors such as banking, healthcare, life sciences and insurance, the Opus platform leverages a 70 billion parameter LWM for complex use cases and an 8 billion parameter model where efficiency is the priority. When evaluated across four key benchmarks, both models outperformed leading large language models (LLMs) including Anthropic’s Claude 3.5, OpenAI-o1 Preview, OpenAI GPT-4o and Google Gemini 1.5 (Flash & Pro). The Opus platform was developed by AppliedAI in partnership with Palantir, McKinsey & Company, and Metalab.
SO WHAT? - According to global research and analyst firm IDC, the sector with the greatest potential for Generative AI with the biggest potential return-on-investment (ROI) is the financial services sector. However, the sector is also one of the most sensitive to risks associated with safety, security and data protection, plus it also has special requirements governing data residency and data transfer. There are a growing number of specific GenAI point solutions being developed, but few horizontal platforms that can help organisations manage a wide range of tasks, with the reliability, security and compliance standards necessary. AppliedAI’s new Opus platform is squarely aimed at regulated industries with sensitive data, equipped with military-grade security and compliance best-practices,
Here are some key details about the new Opus platform:
AppliedAI has launched Opus, a Generative AI platform powered by a first-of-its-kind Large Work Model (LWM) and a proprietary Work Knowledge Graph (WKG). The platform was developed for regulated industries, in partnership with Palantir, McKinsey & Company, and Metalab.
Opus workflows deliver supervised automation outcomes ensuring more consistent, highly auditable and transparent processes in regulated industries such as banking, healthcare, life sciences and insurance.
The LWM generates workflows as structured task sequences and integrates human expert reviews, via two model sizes. The Opus 70 billion parameter model (Alpha 1 Large) is tuned for complex use cases, while the 8 billion parameter model (Alpha 1 Small) is optimised for efficiency.
According to AppliedAI’s testing, both sizes of the Opus Large Work Model outperformed Anthropic’s Claude 3.5, OpenAI-o1 Preview, OpenAI GPT-4o and Google Gemini 1.5 (Flash & Pro) across four benchmark tests: Coverage Ratio (Semantic), Kendall’s Tau (Structural), Dynamic TimeWarping (Structural), and BLEU Score (Semantic).
Meanwhile, the Work Knowledge Graph addresses the complexity and precision needed for unambiguous workflow generation, which is not provided by LLMs alone, nor standard RAG (retrieval-augmented generation) approaches.
The WKG encodes industry-specific procedural and operational knowledge into graph-based representations. Tasks and connections are retrieved using graph attention mechanisms, ensuring workflows are accurate and compliant with domain-specific requirements.
Opus processes multi-modal inputs like text, images, and documents into a unified format. The tokenisation captures semantic, structural, and contextual data to support precise task decomposition.
AppliedAI claims Opus is built with military-grade security architecture, which protects sensitive data, while scaling workflow automation.
In medical coding benchmarks, Opus Alpha 1 Large outperformed leading AI models by 38%, delivering unmatched semantic and structural fidelity.
FUNDING - Founded in 2021, the Applied AI Company (AppliedAI), develops machine learning-driven products that automate mission critical business processes involving manual or repetitive work. With engineering teams in San Francisco, Berlin, and the UAE, the company develops solutions for such sectors such as insurance, pharmaceuticals, healthcare, safety, and government. AppliedAI secured $42 million in funding in 2022, backed by investors including G42, A.R.M Holding, Plug and Play Investment Group, and families investment offices from Europe and the MENA region including Prince Stefan of Liechtenstein.
LINKS
Opus research paper (arXiv)