Patents by Inventor Sanjiv Kumar

Sanjiv Kumar has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).

  • Publication number: 20260330180
    Abstract: Provided are systems and methods that improve the computational efficiency of Transformers or other attention-based neural networks or machine learning models by re-using a number of attention scores between layers and/or heads of the model. To reduce the computational cost of self-attention-based models while achieving comparable or even superior results, example aspects of the present disclosure propose a novel architecture that reuses attention scores computed in one layer in one or multiple subsequent layers.
    Type: Application
    Filed: July 2, 2026
    Publication date: October 22, 2026
    Inventors: Venkata S. Bhojanapalli, Andreas Veit, Ayan Chakrabarti, Frederick Liu, Himanshu Jain, Michal Lukasik, Sanjiv Kumar, Yin-Wen Chang
  • Publication number: 20260329875
    Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing a nearest neighbor search procedure. In one aspect, a method includes receiving a query for one or more nearest neighbor data points to a requested data point from a dataset of multiple data points, maintaining data representing a codebook, where the codebook is associated with a first set and a second set of vector quantization assignments, where the codebook includes multiple centers and each center represents a different partition of the dataset, where the first set of vector quantization assignments assigns each data point to an index of a closest center of the multiple centers and the second set of vector quantization assignments assigns each data point to an index of another center of the plurality of centers, and determining one or more nearest neighbor data points for the query using the data.
    Type: Application
    Filed: May 17, 2024
    Publication date: October 22, 2026
    Inventors: Philip Wenjie Sun, David M. Simcha, David Thomas Dopson, Ruiqi Guo, Sanjiv Kumar
  • Patent number: 12756199
    Abstract: A system includes a flow line, a conduit, a turbine, an energy generator, and a chip. The flow line has an outer circumferential surface and an inner circumferential surface and connected to the well. The conduit is delineated by the inner circumferential surface of the flow line. The turbine is disposed in the conduit of the flow line and configured to convert kinetic energy of a fluid, flowing from the well through the conduit, to rotational movement. The energy generator is connected to the turbine and is configured to convert the rotational movement to electrical energy. The chip is connected to the energy generator and configured to use the electrical energy created by the energy generator to send a signal indicating an operational status of the well to a receiver.
    Type: Grant
    Filed: April 21, 2023
    Date of Patent: October 6, 2026
    Assignee: SAUDI ARABIAN OIL COMPANY
    Inventors: Sanjiv Kumar, Mohammed A. Al-Abdulhay
  • Publication number: 20260300306
    Abstract: A computer-implemented method includes receiving, by a computing system, a query, computing, by the computing system, a similarity matrix based on a first content associated with the query generated by a first transformer and a second content associated with a plurality of documents generated by a second transformer, processing, by the computing system, the similarity matrix via a neural network to generate relevance values respectively corresponding to the plurality of documents, and ranking, by the computing system, the plurality of documents with respect to the query based on the relevance values.
    Type: Application
    Filed: April 24, 2024
    Publication date: October 1, 2026
    Inventors: Himanshu Jain, Ziwei Ji, Andreas Veit, Sashank Jakkam Reddi, Gayan Sadeep Jayasumana Hirimbura Matara Kankanamge, Ankit Singh Rawat, Aditya Krishna Menon, Xinnan Yu, Sanjiv Kumar
  • Publication number: 20260288789
    Abstract: Implementations disclose selecting, in response to receiving a request and from among multiple candidate generative models (e.g., multiple candidate large language models (LLMs)) with differing computational efficiencies, a particular generative model to utilize in generating a response to the request. Those implementations reduce latency and/or conserve computational resource(s) through selection, for various requests, of a more computationally efficient generative model for utilization in lieu of a less computationally efficient generative model. Further, those implementations seek to achieve such benefits, through utilization of more computationally efficient generative models, while also still selectively utilizing less computationally efficient generative models for certain requests to mitigate occurrences of a generated response being inaccurate and/or under-specified.
    Type: Application
    Filed: May 21, 2026
    Publication date: September 24, 2026
    Inventors: Seungyeon Kim, Ankit Singh Rawat, Wittawat Jitkrittum, Hari Narasimhan, Sashank Reddi, Neha Gupta, Srinadh Bhojanapalli, Aditya Menon, Manzil Zaheer, Tal Schuster, Sanjiv Kumar, Toby Boyd, Zhifeng Chen, Emanuel Taropa, Vikram Kasivajhula, Trevor Strohman, Martin Baeuml, Leif Schelin, Yanping Huang
  • Patent number: 12725808
    Abstract: A bipolar plate for a fuel cell, is provided having an anode plate with an anode side and a coolant side, wherein there is formed on the anode side a first structuring in order to form an anode flow field, and a cathode plate with a cathode side and a coolant side, wherein there is formed on the cathode side a second structuring to form a cathode flow field, there being arranged between the anode plate and the cathode plate structural elements to form a coolant flow field, being contacted from the coolant sides of the anode plate and the cathode plate, and having an optimized pressure distribution in a fuel cell stack and an increased stability as compared to the prior art. The structural elements consist of an elastic material. A fuel cell stack and a vehicle including such features are also provided.
    Type: Grant
    Filed: March 20, 2020
    Date of Patent: September 1, 2026
    Assignee: AUDI AG
    Inventors: Adel Jilani, Sanjiv Kumar, Radu P. Bradean, Sebastian Voigt
  • Patent number: 12699874
    Abstract: Provided are systems and methods that improve the computational efficiency of Transformers or other attention-based neural networks or machine learning models by re-using a number of attention scores between layers and/or heads of the model. To reduce the computational cost of self-attention-based models while achieving comparable or even superior results, example aspects of the present disclosure propose a novel architecture that reuses attention scores computed in one layer in one or multiple subsequent layers.
    Type: Grant
    Filed: October 5, 2022
    Date of Patent: August 4, 2026
    Assignee: GOOGLE LLC
    Inventors: Venkata S. Bhojanapalli, Andreas Veit, Ayan Chakrabarti, Frederick Liu, Himanshu Jain, Michal Lukasik, Sanjiv Kumar, Yin-Wen Chang
  • Publication number: 20260220459
    Abstract: Methods and systems for fine-tuning a machine-learning model using a fusion-frame representation are provided. Model weights associated with a layer of a machine-learning model are updated using a projection of a block diagonal matrix into a fusion frame, with each block of the block diagonal matrix projected onto a respective subspace of the fusion frame. This reduces the number of parameters that need to be handled during fine-tuning, reducing demand on processor and memory resources, as it is not necessary to fine-tune all the weights, only the elements of the block diagonal matrix. The use of fusion frames also enables targeting of particular parameters or parameter sets using the subspaces of the fusion frame. Furthermore, embodiments described herein show how the fusion frame implementation lends itself to parallelisation of calculations associated with each respective subspace.
    Type: Application
    Filed: January 30, 2026
    Publication date: July 30, 2026
    Inventors: Sanjiv Kumar, Vikas Singh, Li Zhang, Harshavardhan Adepu
  • Patent number: 12694268
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for multi-stage computationally-efficient inference using a first and second neural network.
    Type: Grant
    Filed: April 14, 2022
    Date of Patent: July 28, 2026
    Assignee: Google LLC
    Inventors: Ankit Singh Rawat, Manzil Zaheer, Aditya Krishna Menon, Sanjiv Kumar, Amr Ahmed
  • Publication number: 20260203674
    Abstract: Provided are systems and methods which more efficiency train embedding models through the use of a cache of item embeddings for candidate items over a number of training iterations. The cached item embeddings can be “stale” embeddings that were generated by a previous version of the model at a previous training iteration. Specifically, at each iteration, the (potentially stale) item embeddings included in the cache can be used when generating similarity scores that are the basis for sampling a number of items to use as negatives in the current training iteration. For example, a Gumbel-Max sampling approach can be used to sample negative items that will enable an approximation of a true gradient. New embeddings can be generated for the sampled negative items and can be used to train the model at the current iteration.
    Type: Application
    Filed: March 12, 2026
    Publication date: July 16, 2026
    Inventors: Erik Michael Lindgren, Ruiqi Guo, Sanjiv Kumar, Sashank Jakkam Reddi
  • Patent number: 12664187
    Abstract: Implementations disclose selecting, in response to receiving a request and from among multiple candidate generative models (e.g., multiple candidate large language models (LLMs)) with differing computational efficiencies, a particular generative model to utilize in generating a response to the request. Those implementations reduce latency and/or conserve computational resource(s) through selection, for various requests, of a more computationally efficient generative model for utilization in lieu of a less computationally efficient generative model. Further, those implementations seek to achieve such benefits, through utilization of more computationally efficient generative models, while also still selectively utilizing less computationally efficient generative models for certain requests to mitigate occurrences of a generated response being inaccurate and/or under-specified.
    Type: Grant
    Filed: June 19, 2023
    Date of Patent: June 23, 2026
    Assignee: GOOGLE LLC
    Inventors: Seungyeon Kim, Ankit Singh Rawat, Wittawat Jitkrittum, Hari Narasimhan, Sashank Reddi, Neha Gupta, Srinadh Bhojanapalli, Aditya Menon, Manzil Zaheer, Tal Schuster, Sanjiv Kumar, Toby Boyd, Zhifeng Chen, Emanuel Taropa, Vikram Kasivajhula, Trevor Strohman, Martin Baeuml, Leif Schelin, Yanping Huang
  • Publication number: 20260170616
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a data item conditioned on a conditioning input. For example, the data item can be audio, video, or an image. The data item is generated across a plurality of denoising steps, with at least some of the denoising steps being performed using a conditional latent random field model.
    Type: Application
    Filed: December 18, 2025
    Publication date: June 18, 2026
    Inventors: Gayan Sadeep Jayasumana Hirimbura Matara Kankanamge, Kanchana Ranasinghe, Srikumar Ramalingam, Andreas Veit, Daniel Glasner, Sanjiv Kumar, Ayan Chakrabarti
  • Patent number: 12619922
    Abstract: Provided are systems and methods which more efficiency train embedding models through the use of a cache of item embeddings for candidate items over a number of training iterations. The cached item embeddings can be “stale” embeddings that were generated by a previous version of the model at a previous training iteration. Specifically, at each iteration, the (potentially stale) item embeddings included in the cache can be used when generating similarity scores that are the basis for sampling a number of items to use as negatives in the current training iteration. For example, a Gumbel-Max sampling approach can be used to sample negative items that will enable an approximation of a true gradient. New embeddings can be generated for the sampled negative items and can be used to train the model at the current iteration.
    Type: Grant
    Filed: November 8, 2022
    Date of Patent: May 5, 2026
    Assignee: GOOGLE LLC
    Inventors: Erik Michael Lindgren, Sashank Jakkam Reddi, Ruiqi Guo, Sanjiv Kumar
  • Publication number: 20260093994
    Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for estimation techniques for efficient neural network inference processing. In some implementations, parameter values for a trained neural network comprising multiple layers are stored, including (i) a matrix of parameter values for at least one layer and (ii) an approximate matrix of values corresponding to the at least one layer. Input is processed using the trained neural network, including determining an input for the at least one layer, computing approximate outputs corresponding to elements in a set using the approximate matrix, and computing intermediate outputs for only a proper subset of the elements in the set using the matrix of parameter values for the at least one layer. The proper subset is determined based on the approximate outputs.
    Type: Application
    Filed: February 14, 2025
    Publication date: April 2, 2026
    Inventors: Yashas Samaga, Varun Yerram, Venkata Sesha Pavana Srinadh Bhojanapalli, Chong You, Sanjiv Kumar, Prateek Jain, Praneeth Kumar Netrapalli
  • Publication number: 20260093967
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for increasing sparsity to improve neural network efficiency. In some implementations, a system stores parameter values of parameter matrices of one or more layers of a neural network. The parameter values of the parameter matrices include (i) weight values of the one or more layers of the neural network, and (ii) predictor values that have been trained to predict levels of importance of items processed by the neural network. The system generates an output, including: determining a value for each of multiple items using the predictor values, selecting a proper subset of the items based on the values in the vector based on a threshold, and generating output of the one or more layers limiting computation based on the selected proper subset.
    Type: Application
    Filed: October 2, 2025
    Publication date: April 2, 2026
    Inventors: David Ethan Culler, Prateek Jain, Zhipeng Jia, Sanjiv Kumar, Jeremiah Willcock, Chong You, Shreya Pathak, Lin Chen, Xinnan Yu, Venkata Sesha Pavana Srinadh Bhojanapalli, Suvinay Subramanian, Felix Ren-Chyan Chern, Alek Alexandrov Andreev, Praneeth Kumar Netrapalli, Kan Wu, Henry Marc Levy
  • Patent number: 12579439
    Abstract: A method includes receiving, by a computing device, training data to train a neural network, wherein the training data comprises a plurality of inputs and a plurality of corresponding labels. The method also includes mapping, by a representation learner of the neural network, the plurality of inputs to a plurality of feature vectors. The method additionally includes training a kernelized classification layer of the neural network to perform nonlinear classification of an input feature vector into one of a plurality of classes, wherein the kernelized classification layer is based on a kernel which enables the nonlinear classification, and wherein the kernel is selected from a space of positive definite kernels based on application of a nonlinear softmax loss function to the plurality of feature vectors and the plurality of corresponding labels. The method further includes outputting a trained neural network comprising the representation learner and the trained kernelized classification layer.
    Type: Grant
    Filed: April 30, 2021
    Date of Patent: March 17, 2026
    Assignee: Google LLC
    Inventors: Gayan Sadeep Jayasumana Hirimbura Matara Kankanamge, Srikumar Ramalingam, Sanjiv Kumar
  • Patent number: 12561583
    Abstract: Generally, the present disclosure provides systems and methods for performing machine learning in hyperbolic space. Specifically, techniques are provided which enable the learning of a classifier (e.g., large-margin classifier) for data defined within a hyperbolic space (e.g., which may be particularly beneficial for data possessing a hierarchical structure).
    Type: Grant
    Filed: April 12, 2021
    Date of Patent: February 24, 2026
    Assignee: GOOGLE LLC
    Inventors: Ankit Singh Rawat, Manzil Zaheer, Aditya Krishna Menon, Sanjiv Kumar, Melanie Weber
  • Patent number: 12547759
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for privacy preserving training of a machine learning model.
    Type: Grant
    Filed: August 14, 2020
    Date of Patent: February 10, 2026
    Assignee: Google LLC
    Inventors: Ananda Theertha Suresh, Xinnan Yu, Sanjiv Kumar, Sashank Jakkam Reddi, Venkatadheeraj Pichapati
  • Publication number: 20260037593
    Abstract: A computing system can obtain a machine-learned model comprising one or more parameters having a four-bit binary format. The four-bit binary format can correlate a plurality of sixteen respective binary values to a plurality of sixteen corresponding numerical values represented by the binary values. The sixteen numerical values can be symmetric about a median. A plurality of step sizes between the sixteen numerical values can be non-uniform. The computing system can obtain one or more input values for one or more layers of the machine-learned model. The computing system can process, based at least in part on the one or more parameters having the four-bit binary format, the one or more input values to generate one or more output values.
    Type: Application
    Filed: August 2, 2024
    Publication date: February 5, 2026
    Inventors: Jian Li, Sanjiv Kumar, Felix Chern, Zhifeng Chen
  • Patent number: 12530576
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for accounting for long-tail training data.
    Type: Grant
    Filed: July 14, 2021
    Date of Patent: January 20, 2026
    Assignee: Google LLC
    Inventors: Aditya Krishna Menon, Sanjiv Kumar, Himanshu Jain, Andreas Veit, Ankit Singh Rawat, Gayan Sadeep Jayasumana Hirimbura Matara Kankanamge