Neurond, one of the top AI companies in Vietnam, has marked a dynamic partnership with Intel, a global leader in hardware innovation.
This collaboration leverages Intel’s AI-ready technology to elevate one of our AI solutions, the Dr. Parser tool. It has helped deliver unparalleled speed and efficiency to the tool performance while showcasing the transformative potential of Intel’s hardware in real-world AI applications.
Partnership Based on Innovation
The Challenge: Speeding Up Dr. Parser
Neurond’s Dr. Parser is an AI-powered tool designed to streamline the hiring process. It helps automate the resume and CV information extraction process, which saves recruiters valuable time and improves the candidate experience.
However, the complex deep-learning tasks involved, particularly the Optical Character Recognition (OCR) required for document extraction from images, demanded significant processing power.
Before the optimization, processing a typical 3-4 page resume could take approximately 20 seconds on standard CPU hardware, with the OCR module consuming the majority of this time.
The Solution: Intel Hardware and OpenVINO™
The partnership with Intel changed the game. Neurond’s team focused on optimizing the most time-intensive component: the OCR module. The process involved:
- Converting the existing PyTorch AI model to the ONNX (Open Neural Network Exchange) format.
- Implementing Intel’s OpenVINO™ toolkit for inference, enabling optimized performance across different Intel hardware types (CPU, GPU).
By integrating Intel’s Core Ultra 7 258V processor, Intel Arc 140V graphics, and the OpenVINO toolkit, Neurond achieved dramatic performance gains. Document processing times dropped sharply, with extraction tasks now completing more quickly than the original time, often under 5 seconds for a 2-page document. These results, validated across a dataset of 257 resumes, highlight the power of Intel’s hardware to accelerate AI workloads.
The Results: A Significant Improvement in Performance
The results were remarkable. By running the optimized Dr. Parser on Intel’s hardware, particularly utilizing the integrated GPU and the OpenVINO™ toolkit, Neurond achieved a significant reduction in document processing time.
Benchmark tests showed substantial decreases in the average latency for both individual page processing within the OCR module and the overall document extraction process. Even when running optimized models on the CPU using OpenVINO™, notable performance improvements were observed compared to the original PyTorch configuration.
Mutual Benefits, Shared Vision
This partnership offers significant advantages for both companies.
For Neurond, this collaboration results in faster and more scalable Dr. Parser; and it’s ready to meet the demands of high-volume hiring environments. Recruiters benefit from reduced screening times, while candidates enjoy a smoother, more responsive application process. By integrating with Intel’s hardware, we solidify Dr. Parser’s position as a leader in AI-driven recruitment solutions, pushing the boundaries of what resume parsing can achieve.
Intel, in turn, proves the real-world impact of its AI-ready hardware and OpenVINO toolkit. Our success with Dr. Parser demonstrates the performance, efficiency, and versatility of Intel’s Core Ultra processors and Arc graphics, positioning Intel as a go-to choice for AI workloads in enterprise applications. This partnership also highlights Intel’s commitment to supporting innovative companies like Neurond, fostering advancements that redefine industry standards.
Looking Ahead
The Neurond-Intel partnership is just the beginning. With Intel’s dedicated support, including access to virtual CPUs and expert technical teams, Neurond is poised to further refine Dr. Parser’s capabilities. Future optimizations, including potential NPU integration, promise even greater efficiency and power savings. Together, Neurond and Intel are paving the way for a new era of intelligent, high-speed information extraction technology.
Stay tuned for more updates on the Neurond and Intel collaboration.
