• last updated : 04 July, 2024

XLSCOUT and Hugging Face Introduce ParaEmbed 2.0: Powerful Embedding Model Tailored for Patents and IP

Category: Press Release
ParaEmbed 2.0

XLSCOUT, a pioneer in AI-powered solutions for intellectual property based in Toronto, proudly announces the release of ParaEmbed 2.0, a sophisticated proprietary embedding model.

This release marks a significant milestone in our partnership with Hugging Face’s Expert Support Program.

Together, we are focusing on leveraging cutting-edge AI and open-source technologies to profoundly enhance the analysis and interpretation of complex patent documents, including specialized patent terminology, context, and interrelationships.

ParaEmbed 2.0, fine-tuned on a diverse, high-quality set of patent data curated by domain experts, achieves a remarkable 23% boost in accuracy over its predecessor, ParaEmbed 1.0v, launched in October 2023.

This improvement enables more precise context capture and comparison of patents against prior art and other standards, enhancing the quality and reliability of patent analysis.

Evolving Patent Analysis

Initially, XLSCOUT utilized proprietary AI models for patent analytics but faced challenges with the nuanced context of technical and specialized patent claims using closed-source models like GPT-4 and text-embedding-ada-002.

Transitioning to open-source models, including BGE-base-v1.5, Llama 2 70B, Falcon 40B, and Mixtral 8x7B, and refining these with our proprietary patent data under Hugging Face’s guidance, has facilitated the creation of more customized and effective solutions.

This transition has revolutionized the precision in understanding technical documents and patents.

Deepening Collaboration with Hugging Face

This collaboration has been instrumental in elevating the performance and quality of XLSCOUT’s offerings:

1. Initial Development

Constructing a customized TorchServe inference server on Google Cloud Platform (GCP) with Distributed Data Parallel (DPP) technology allowed serving multiple replicas efficiently. Integrating ONNX optimizations further enhanced our capability to process approximately 300 embeddings per second.

2. Enhanced Performance via Fine-Tuning

The fine-tuning process using expert-curated patent data not only refined the model’s contextual accuracy but also significantly boosted performance metrics, ensuring more precise detection of relevant prior art.

3. High Throughput Serving

Leveraging Hugging Face’s Inference Endpoints, equipped with built-in load balancing, XLSCOUT now efficiently delivers around 2700 embeddings per second in production settings.

4. LLM Prompting and Inference

Our collaboration also advanced the model’s ability to generate accurate and context-specific patent drafts. Instructional data formatting and fine-tuning with models from various leaders in the field have refined the drafting process to produce even more accurate outputs.

XLSCOUT’s LLMs & Generative AI-Powered IP Solutions

XLSCOUT continues to lead in providing advanced AI-powered solutions, enhancing patent-related workflows for corporations, universities, and law firms. Our solutions, including the Novelty Checker LLM, Invalidator LLM, and Drafting LLM, expedite and refine processes from novelty searches and invalidation studies to comprehensive patent drafting.

A Partnership Driving Innovation

“We are thrilled to collaborate with Hugging Face,” stated Mr. Sandeep Agarwal, CEO of XLSCOUT. ” This partnership combines the unparalleled capabilities of Hugging Face’s open-source models, tools, and team with our deep expertise in patents. By fine-tuning these models with our proprietary data, we are poised to revolutionize how patents are drafted, analyzed, and licensed.”

This collaboration focuses on training open-source models using XLSCOUT’s extensive patent data, merging specialized knowledge with advanced AI to optimize models for patent research. The results are enabling more informed decisions and insights into complex patent documents.

Commitment to Ongoing Innovation

As innovators in the application of AI to intellectual property, XLSCOUT is dedicated to pushing the boundaries of AI-driven innovation in IP.

Our partnership with Hugging Face is a pivotal step in aligning cutting-edge AI with practical IP analysis applications.

XLSCOUT and Hugging Face continue to redefine standards in patent analysis, fostering innovation and shaping the future of intellectual property management. We’re excited to continue this awesome journey together!

About Hugging Face

Hugging Face leads the transformation of how natural language processing (NLP), automated reasoning, and generative AI models are developed and implemented in today’s technology landscape.

Renowned for its vibrant, open-source ecosystem, Hugging Face fosters innovation in AI through collaborative contributions from a global community of researchers and developers. This collaboration fuels advances in AI technology, making state-of-the-art models more accessible and efficient across a diverse range of industries and applications.


XLSCOUT is at the forefront of the global innovation and IP industry, harnessing the potential of advanced AI technologies like LLMs to optimize ideation, create high-value patents, and monetize innovation.

The company combines its deep patent domain knowledge and industry experience with cutting-edge AI technologies to develop the most accurate, comprehensive, and intelligent patent database.