Andrew Ng Patents & Innovation Profile — PatSnap Eureka
Andrew Y. Ng: Patent Portfolio & Innovation Analysis
Andrew Y. Ng is one of the most influential figures in artificial intelligence, holding 77 curated patents across deep learning, speech recognition, and computer vision, with filings spanning 1995 to 2025. His portfolio is primarily assigned to LandingAI Inc. and covers foundational methods in neural network training, end-to-end speech recognition, and industrial visual inspection — with a broader dataset of 1,714 records attributed to variants of his name across global patent databases.
Patent Filing Activity
Peak filing year was 2023 with 10 filings; the 2019–2025 LandingAI period is the most prolific phase of the portfolio.
Andrew Y. Ng's Patent Filing Patterns
Three distinct career phases are visible in the filing data: an early sparse period, a Baidu-driven acceleration peaking in 2015, and a prolific LandingAI era peaking in 2023 with 10 filings.
Annual Patent Filings
Peak year 2023 with 10 filings; 2019 marked the LandingAI founding surge with 9 filings. 7 filings already recorded in 2025.
Technology Domain Breakdown
Neural networks (G06N3) and speech recognition (G10L15) together represent 48% of the curated portfolio, with computer vision and MLOps each at 18%.
Andrew Y. Ng's Core Areas of Innovation
Andrew Y. Ng's patent portfolio spans five technology domains, tracing a career arc from enterprise distributed systems through speech AI and deep learning infrastructure to industrial computer vision and data-centric MLOps.
Neural Networks & Deep Learning Infrastructure
8 patentsThis is the portfolio's core technical domain, covering architecture and training of deep neural networks. Patents address real-time neural text-to-speech, adaptive training dataset refinement through deployment feedback, and guided deep learning error analysis workflows — perennial challenges at the heart of practical AI deployment.
- Systems and methods for real-time neural text-to-speech (US20180247636A1)
- Model Management System for Improving Training Data Through Machine Learning Deployment (US20220300855A1)
- Guided workflow for deep learning error analysis (US12333792B2)
Speech Recognition & Transcription
8 patentsA closely clustered group of patents arising from Andrew Ng's leadership of Baidu's AI research division covers end-to-end deep learning approaches to speech recognition, anchored by the Deep Speech system. These filings replaced traditional pipeline-based ASR architectures with recurrent neural networks trained on large datasets, demonstrating that simpler models trained with sufficient data outperform complex hand-engineered systems.
- Systems and methods for speech transcription (US20160171974A1)
- Deep learning models for speech recognition (US11562733B2)
- Systems and methods for speech transcription — EP grant (EP3180785B1)
Computer Vision & Visual Inspection
6 patentsPatents in this domain reflect LandingAI Inc.'s commercial focus: applying machine learning to manufacturing quality control and agricultural computer vision. Inventions address defect detection, impurity identification, partial labelling for efficient training data creation, and data-centric mislabel detection — extending deep learning theory into industrial hardware systems.
- AI-optimized harvester configured to maximize yield and minimize impurities (US11412657B2)
- Partial labeling mechanism for quick and accurate training of machine learning models (US12579793B1)
- Data centric mislabel detection (US12482242B1)
Machine Learning Systems & MLOps
6 patentsA cluster of patents addresses operational and lifecycle challenges of deploying machine learning models in real environments, covering user-generated visual guides for training data classification, integrated ML and rules platforms for improved accuracy, and radiology image classification from noisy images. These inventions collectively represent the emerging "data-centric AI" philosophy.
- User-generated visual guide for the classification of images (US11182646B2)
- Systems and methods for radiology image classification from noisy images (US11798159B2)
Distributed Computation & Task Management
5 patentsAn earlier cluster, dating from the early 2000s and assigned to SAP Aktiengesellschaft, covers distributed systems for task management and fault recovery in data processing environments. These patents predate Andrew Ng's primary AI research career and reflect earlier work in enterprise systems engineering. The key patent has accumulated 82 citations, indicating sustained relevance to the systems management field.
- Managing tasks in a data processing environment (US20040226013A1)
Andrew Y. Ng's Highest-Impact IP
The most cited patent — Baidu's speech transcription filing — has accumulated 159 citations, reflecting the Deep Speech programme's documented impact on the ASR field and subsequent voice interface development.
| Patent Number | Title | Year | Citations | Assignee | Status |
|---|---|---|---|---|---|
| US20160171974A1 | Systems and methods for speech transcription | 2015 | 159 ↑ | BAIDU USA LLC | active |
| US7801645B2 | Robotic vacuum cleaner with edge and object detection system | 2004 | 153 ↑ | SHARPER IMAGE ACQUISITION LLC | inactive |
| US20200211154A1 | Method and system for privacy-preserving fall detection | 2019 | 94 ↑ | ALTUMVIEW SYSTEMS INC. | active |
| US20040226013A1 | Managing tasks in a data processing environment | 2003 | 82 ↑ | SAP AKTIENGESELLSCHAFT | active |
| US20080137989A1 | Arrangement and method for three-dimensional depth image construction | 2007 | 74 ↑ | SAXENA, ASHUTOSH | inactive |
| US20140304335A1 | Systems and methods for interactive experiences and controllers therefor | 2014 | 65 ↑ | TIMEPLAY ENTERTAINMENT CORPORATION | inactive |
| US20180247636A1 | Systems and methods for real-time neural text-to-speech | 2018 | 64 ↑ | BAIDU USA, LLC | active |
| US20160055261A1 | User-controlled graph analysis system | 2014 | 51 ↑ | CRAY INC. | active |
Andrew Y. Ng's Research Collaborators
Most Frequent Co-Inventors
Collaboration Highlights
Andrew Ng's collaboration network reveals two distinct phases: a Baidu-era team anchored by Gregory Diamos and Sanjeev Satheesh working on the Deep Speech ASR programme, and a near-complete personnel transition to a LandingAI team centred on Kai Yang and Yu Qing Zhou for industrial computer vision and data-centric AI tooling. The absence of overlap between these two groups signals how completely Andrew Ng rebuilt his research organisation between his Baidu departure in 2017 and LandingAI's scaling from 2019 onward.
- Gregory Diamos 14–15 joint patents
- Kai Yang 13 joint patents
- Yu Qing Zhou 12 joint patents
- Sanjeev Satheesh 12 joint patents
Research Literature by Andrew Y. Ng
1,841 papers indexed · Research themes span crowdsourcing methodology, medical AI imaging, and computer vision scene understanding — each closely coupled to patent filings by 12–24 months.
| Title | Year | Citations | Affiliation / Source |
|---|---|---|---|
| Evaluating non-expert annotations for natural language tasks | 2008 | 1,510 ↑ | Stanford University / Dolores Labs |
| Deep learning for chest radiograph diagnosis: CheXNeXt | 2018 | 1,019 ↑ | Stanford University / Duke University |
| Grounded Compositional Semantics for Finding and Describing Images with Sentences | 2014 | 725 ↑ | Stanford University / Google Inc. |
| Deep-learning-assisted diagnosis for knee MRI: MRNet | 2018 | 540 ↑ | Stanford University |
| 3-D Depth Reconstruction from a Single Still Image | 2007 | 529 ↑ | Stanford University |
Crowdsourcing & Data Annotation
The landmark 2008 study on non-expert annotations via Amazon Mechanical Turk (1,510 citations) is among the most cited in NLP methodology and directly prefigures the data-centric AI philosophy embedded in Andrew Ng's LandingAI patents on labelling, error analysis, and mislabel detection.
Medical AI & Clinical Imaging
The CheXNeXt chest radiograph algorithm (1,019 citations) and MRNet for knee MRI interpretation (540 citations), produced through Stanford's Center for AI in Medicine and Imaging, have directly influenced how AI is evaluated for clinical deployment and generated patent-proximate inventions including US11798159B2.
Computer Vision & Scene Understanding
The 3D depth reconstruction from single images work (529 citations) maps directly to Stanford-assigned depth image construction patents in the portfolio. The grounded compositional semantics paper (725 citations) bridges vision and language understanding — a precursor to modern multimodal AI systems.
Patent Jurisdictions
Andrew Y. Ng's portfolio spans 8 jurisdictions, with the US-WO-EP-KR-CN axis for the Baidu speech AI family reflecting the global commercial importance of voice technology in those markets.
Filing Markets
The United States accounts for 46 of the 77 curated patents, reflecting the US-centric operations of Stanford, Baidu USA LLC, and LandingAI Inc. The PCT and EP filings (9 each) are concentrated in the Baidu speech AI family, consistent with the global voice assistant and telecommunications market. More recent LandingAI filings concentrate heavily in the US, consistent with a startup focused on domestic industrial clients maintaining a leaner international prosecution budget.
Why Andrew Y. Ng's Portfolio Matters
Strategic implications for patent attorneys, in-house IP teams, and R&D strategists working in AI-adjacent fields.
FTO Considerations
The speech recognition and neural text-to-speech clusters held by Baidu USA LLC remain active in the US and EP. Developers building voice interfaces, ASR pipelines, or neural audio synthesis products should assess these patent families carefully. The Deep Speech approach — end-to-end RNN-based ASR without phoneme dictionaries — is broad enough to warrant thorough claim mapping before commercialisation. LandingAI Inc.'s growing cluster covering training data management, visual guide-based labelling, error analysis workflows, partial labelling mechanisms, and mislabel detection constitutes a coherent and defensible IP position in the data-centric AI space.
Prior Art Relevance
The most cited patent — US20160171974A1 (159 citations) — is substantive prior art for any ASR, voice interface, or neural audio system. The 3D depth image construction patent (74 citations) from Stanford has been cited by robotics, autonomous vehicle, and augmented reality developers. The distributed task management patent from 2003 (82 citations) retains relevance as prior art decades after filing. Inventor name disambiguation is critical: standard searches for "Andrew Ng" will over-include — jurisdiction filtering and assignee cross-referencing against Baidu USA LLC, LandingAI Inc., and Stanford is recommended.
Frequently Asked Questions About Andrew Y. Ng's Patents
Analyse Andrew Y. Ng's Full Patent Landscape in PatSnap Eureka IP
Access inventor disambiguation tools, claim-level analysis, citation network mapping, and competitive landscape assessments across deep learning, speech AI, and industrial computer vision.
References & Patent Sources
- US20160171974A1 — Systems and methods for speech transcription (Baidu USA LLC, 2015) — PatSnap Eureka
- US20180247636A1 — Systems and methods for real-time neural text-to-speech (Baidu USA LLC, 2018) — PatSnap Eureka
- US20040226013A1 — Managing tasks in a data processing environment (SAP Aktiengesellschaft, 2003) — PatSnap Eureka
- US20080137989A1 — Arrangement and method for three-dimensional depth image construction (Stanford, 2007) — PatSnap Eureka
- USPTO Patent Full-Text and Image Database — United States Patent and Trademark Office
- Espacenet Patent Search — European Patent Office
- PATENTSCOPE — World Intellectual Property Organisation (WIPO)
- PatSnap Innovation Intelligence Platform — patsnap.com
PatSnap Eureka searches 208M+ patents and papers to answer instantly.