logo

For the Inventor, By the Inventor

Petach Tikva, Israel

Shai Ardazi

This inventor holds 22 USPTO granted patents and 15 published patent applications, primarily in Language Model Optimization. Top assignee: Intuit, Inc.. Active years: 2026.

IDiyas Innovation Intelligence. (2026). Inventor Profile: Shai Ardazi. Retrieved from https://idiyas.com/inventor/shai-ardazi

Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile

USPTO Granted Patents = 22 

% Patents Active = 100.0

Average Co-Inventor Count = 4.0

ph-index = 1


Company Filing History:

goldMedal22 out of 2,549 
 
Intuit, Inc.
 patents

Years Active: 2026

Loading Chart...
Areas of Expertise:
Speculative Decoding
Language Model
Policy Optimization
Caching Techniques
Entropy-Based Decoding
Generative AI
Retrieval-Augmented Generation
Machine Learning
Token Prediction
Hallucination Detection
Computational Efficiency
22 patents (USPTO):Explore Patents
1. 12755209 - Advantage weighted sampling for group-based relative policy optimization in machine learning (2026-10-06),
2. 12755110 - Taint-aware, entropy-based set block decoding for language model-based token prediction (2026-10-06),
3. 12754958 - Template guided speculative decoding (2026-10-06),
4. 12737551 - Computer-aware caching for low latency LLM serving (2026-09-15),
5. 12737544 - Unsupervised detection of hallucinations in language model outputs (2026-09-15),
6. 12725089 - Speculative retrieval proxy for retrieval augmented generation (2026-09-01),
7. 12724979 - Closed-loop speculative decoding with real-time prompt patching for language model output generation (2026-09-01),
8. 12718073 - Hierarchical generative AI system and method utilizing logit-based reasoning transfer between models (2026-08-25),
9. 12712733 - PARITY TOKEN FOR ERROR-LOCALIZED BLOCK DECODING (2026-08-18),
10. 12705462 - Decoder architecture for early-stop automated retrieval augmented generation (2026-08-11),
11. 12699841 - Importance aware selective correction for speculative decoding in large models (2026-08-04),
12. 12694213 - Entropy-based set block decoding utilizing an adaptive entropy threshold (2026-07-28),
13. 12670406 - Confidence-based reward for group relative policy optimization in language models (2026-06-30),
14. 12670326 - Pivotal token search by perplexity (2026-06-30),
15. 12664438 - Computationally efficient reasoning language models (2026-06-23),

Title: Shai Ardazi: Innovator in Model Merging Technologies

Introduction

Shai Ardazi is a prominent inventor based in Petach Tikva, Israel. He has made significant contributions to the field of model merging, particularly in the context of machine learning and artificial intelligence. With a total of 22 patents to his name, Ardazi is recognized for his innovative approaches to enhancing model performance across multiple domains.

Latest Patents

Ardazi's latest patents include groundbreaking techniques for model merging. One of his patents, titled "Gradient-based model merging," outlines a method that processes multiple questions to generate predicted answers. This method involves generating gradient vectors for each model to minimize loss values, ultimately leading to a single merged model that effectively integrates knowledge from various domains. Another notable patent, "Model merging via Riemannian barycenters of high-dimensional transformer weights," describes a method for merging transformer models by optimizing weight matrices on a curved manifold. This innovative approach allows for the creation of a merged transformer model that combines the strengths of multiple models.

Career Highlights

Shai Ardazi currently works at Intuit, Inc., where he applies his expertise in machine learning to develop advanced technologies. His work focuses on improving the efficiency and accuracy of models used in various applications. Ardazi's contributions have positioned him as a key player in the field of artificial intelligence.

Collaborations

Throughout his career, Ardazi has collaborated with talented individuals such as Linoy Cohen and Guy Lev. These partnerships have fostered a creative environment that encourages innovation and the development of cutting-edge technologies.

Conclusion

Shai Ardazi is a distinguished inventor whose work in model merging is paving the way for advancements in artificial intelligence. His patents reflect a deep understanding of complex algorithms and their applications in real-world scenarios. As he continues to innovate, Ardazi's contributions will undoubtedly shape the future of machine learning technologies.

Profile summary based on public USPTO records.
Data Sources: USPTO Patent Grant XML, Patent Center, EPO & CIPO • Normalized by IDiyas Innovation Graph. Methodology & provenance architecturePlease report any incorrect information to [email protected]
Loading…