Computer Vision Expert

About PatSnapPatsnap empowers IP and R&D teams by providing better answers, so they can make faster decisions with more confidence. Founded in 2007, Patsnap is the global leader in AI-powered IP and R&D intelligence. Our domain-specific LLM, trained on our extensive proprietary innovation data, coupled with Hiro, our AI assistant, delivers actionable insights that increase productivity for IP tasks by 75% and reduce R&D wastage by 25%. IP and R&D teams collaborate better with a user-friendly platform across the entire innovation lifecycle. Over 15,000 companies trust Patsnap to innovate faster with AI, including NASA, Tesla, PayPal, Sanofi, Dow Chemical, and Wilson Sonsini.

Key Responsibilities

• Drive innovation and achieve core objectives in OCR, image retrieval, and related computer vision domains.
• Explore and implement multimodal technologies across existing projects, identifying new applications and opportunities.
• Lead research and development efforts in advanced computer vision algorithms and techniques.
• Collaborate with cross-functional teams to integrate computer vision solutions into product offerings.

Desired Qualifications

• Master’s degree in Computer Science, Software Engineering, Electronic Engineering, Mathematics, Statistics, or a related field; Ph.D. strongly preferred.
• Minimum of 2 years of hands-on experience in developing and implementing computer vision algorithms.
• Extensive research and practical experience in text OCR, table OCR, and image retrieval, with a deep understanding of state-of-the-art technologies.
• Strong foundation in multimodal technologies, including experience with Vision-Language Models (VLMs) and familiarity with multimodal architectures such as CLIP and LLaVA.
• Demonstrated ability to integrate pre-trained models for document QA, table extraction, image retrieval, and related tasks.
• Passion for cutting-edge technologies and a track record of successful product delivery.
• Publications in peer-reviewed journals or top-tier conferences are preferred.