Inventor · disambiguated record
Harshit Khaitan
Also filed as: KHAITAN HARSHIT
32 granted patents·8 pending applications·493 citations·filing 2016–2025
97Inventor score
Top patents by PatentIndex Score
40 records- 0199US9836691B1Neural network instruction set architectureGOOGLE INC·Filed 2017·Granted Dec 5, 2017·108 cites·17 claims
- 0299US9710265B1Neural network compute tileGOOGLE INC·Filed 2017·Granted Jul 18, 2017·201 cites·17 claims
- 0398US10175980B2Neural network compute tileGOOGLE LLC·Filed 2016·Granted Jan 8, 2019·38 cites·20 claims
- 0497US11709783B1Tensor data distribution using grid direct-memory access (DMA) controllerMETA PLATFORMS INC·Filed 2020·Granted Jul 25, 2023·14 cites·20 claims
- 0596US10504022B2Neural network accelerator with parameters resident on chipGOOGLE LLC·Filed 2018·Granted Dec 10, 2019·22 cites·20 claims
- 0696US9959498B1Neural network instruction set architectureGOOGLE LLC·Filed 2016·Granted May 1, 2018·21 cites·16 claims
- 0795US11467675B1Multi-component detection of gesturesFACEBOOK TECH LLC·Filed 2021·Granted Oct 11, 2022·9 cites·21 claims
- 0894US10496326B2Hardware double buffering using a special purpose computational unitGOOGLE LLC·Filed 2019·Granted Dec 3, 2019·9 cites·17 claims
- 0994US9946539B1Accessing data in multi-dimensional tensors using addersGOOGLE LLC·Filed 2017·Granted Apr 17, 2018·19 cites·20 claims
- 1093US10534607B2Accessing data in multi-dimensional tensors using addersGOOGLE LLC·Filed 2018·Granted Jan 14, 2020·10 cites·12 claims
- 1193US10248908B2Alternative loop limits for accessing data in multi-dimensional tensorsGOOGLE LLC·Filed 2017·Granted Apr 2, 2019·10 cites·19 claims
- 1292US12001893B1Distributed synchronization schemeMETA PLATFORMS INC·Filed 2020·Granted Jun 4, 2024·4 cites·20 claims
- 1392US11704562B1Architecture for virtual instructionsMETA PLATFORMS INC·Filed 2020·Granted Jul 18, 2023·4 cites·20 claims
- 1491US12061968B2Neural network instruction set architectureGOOGLE LLC·Filed 2022·Granted Aug 13, 2024·1 cites·21 claims
- 1591US10108538B1Accessing prologue and epilogue dataGOOGLE LLC·Filed 2017·Granted Oct 23, 2018·7 cites·20 claims
- 1690US10802956B2Accessing prologue and epilogue dataGOOGLE LLC·Filed 2018·Granted Oct 13, 2020·8 cites·20 claims
- 1785US2025045559A1Neural network instruction set architectureGOOGLE LLC·Filed 2024·Application pending·0 cites
- 1883US12455737B2Neural network compute tileGOOGLE LLC·Filed 2023·Granted Oct 28, 2025·0 cites·16 claims
- 1981US11422801B2Neural network compute tileGOOGLE LLC·Filed 2019·Granted Aug 23, 2022·2 cites·20 claims
- 2080US10175912B1Hardware double buffering using a special purpose computational unitGOOGLE LLC·Filed 2017·Granted Jan 8, 2019·2 cites·17 claims
- 2180US2024078417A1Neural network accelerator with parameters resident on chipGOOGLE LLC·Filed 2023·Application pending·0 cites
- 2279US11816480B2Neural network compute tileGOOGLE LLC·Filed 2022·Granted Nov 14, 2023·0 cites·16 claims
- 2379US11727259B2Neural network accelerator with parameters resident on chipGOOGLE LLC·Filed 2022·Granted Aug 15, 2023·0 cites·11 claims
- 2478US2025231765A1Neural network compute tileGOOGLE LLC·Filed 2025·Application pending·0 cites
- 2578US2025265444A1Neural network instruction set architectureGOOGLE LLC·Filed 2025·Application pending·0 cites
- 2677US2025232164A1Neural network accelerator with parameters resident on chipGOOGLE LLC·Filed 2025·Application pending·0 cites
- 2776US11176493B2Virtualizing external memory as local to a machine learning acceleratorGOOGLE LLC·Filed 2019·Granted Nov 16, 2021·2 cites·20 claims
- 2873US11379707B2Neural network instruction set architectureGOOGLE LLC·Filed 2017·Granted Jul 5, 2022·1 cites·20 claims
- 2972US2025232223A1Virtualizing external memory as local to a machine learning acceleratorGOOGLE LLC·Filed 2025·Application pending·0 cites
- 3070US12417406B2Virtualizing external memory as local to a machine learning acceleratorGOOGLE LLC·Filed 2021·Granted Sep 16, 2025·0 cites·20 claims
- 3169US11893159B2Multi-component detection of gesturesMETA PLATFORMS TECH LLC·Filed 2022·Granted Feb 6, 2024·0 cites·20 claims
- 3269US10885434B2Alternative loop limits for accessing data in multi-dimensional tensorsGOOGLE LLC·Filed 2019·Granted Jan 5, 2021·1 cites·20 claims
- 3363US11501144B2Neural network accelerator with parameters resident on chipGOOGLE LLC·Filed 2019·Granted Nov 15, 2022·0 cites·8 claims
- 3462US11099772B2Hardware double buffering using a special purpose computational unitGOOGLE LLC·Filed 2019·Granted Aug 24, 2021·0 cites·14 claims
- 3558US12265492B2Circular buffer for input and output of tensor computationsMETA PLATFORMS INC·Filed 2023·Granted Apr 1, 2025·0 cites·20 claims
- 3658US2024143525A1Transferring non-contiguous blocks of data using instruction-based direct-memory access (dma)META PLATFORMS INC·Filed 2022·Application pending·0 cites
- 3755US12197362B2Batch matrix multiplication operations in a machine learning acceleratorMETA PLATFORMS INC·Filed 2023·Granted Jan 14, 2025·0 cites·20 claims
- 3855US11954580B2Spatial tiling of compute arrays with shared controlMETA PLATFORMS INC·Filed 2020·Granted Apr 9, 2024·0 cites·18 claims
- 3955US11922306B2Tensor controller architectureMETA PLATFORMS INC·Filed 2020·Granted Mar 5, 2024·0 cites·18 claims
- 4051US2022092408A1Neural network weight distribution using a tree direct-memory access (dma) busFACEBOOK INC·Filed 2020·Application pending·0 cites
Join the waitlist — get patent alerts
Get an alert when Harshit Khaitan files or is granted a new patent.
We store only your email — no account needed. See our privacy policy.
Identity basis: PatentsView inventor disambiguation (2025Q4-odp release). How scoring works →