Abstract: A system determines whether a first operation performed by an MLM is compatible with one of a first encryption scheme and a second encryption scheme, wherein the MLM is distributed over at least one client device and at least one server. In response to determining that the first operation is compatible with the first encryption scheme, the system: encrypts data associated with the first operation using the first encryption scheme; and transmits the data encrypted by the first encryption scheme to the at least one server configured to apply the first operation. In response to determining that the first operation is incompatible with the first encryption scheme, the system: encrypts the data associated with the first operation using the second encryption scheme; and transmits the data encrypted by the second encryption scheme to the at least one server configured to apply the first operation.

Abstract: Aspects of the present disclosure include a method for verifying live user presence in an online session, comprising receiving first and second video streams capturing a user during the session from a first position and a different second position, respectively. The method further comprises initiating a challenge by providing, for presentation on a different second display of a second device, a code, and providing, for presentation on a first display, an instruction to the user to respond to the challenge using the second device. The method further comprises detecting, based on at least one of the video streams, physical actions of the user during the challenge, receiving, from the second device, user input events representing a user response to the challenge, and determining, based on at least one of the physical actions or the user input events, whether the user successfully completed the challenge.

Abstract: A method and system is disclosed for generating video content where speech audio is accurately synchronized with facial movements. The method begins by converting input text into speech audio using a text-to-speech (TTS) system. An original video is then processed to identify and extract the mouth area of the human face. The method generates a plurality of frames of the mouth area corresponding to segments of the speech audio using a trained generative neural network. These frames are lip-synced with the corresponding audio segments. The extracted mouth area frames are overlaid onto the original video, followed by performing facial reconstruction using a trained facial reconstruction model to correct distortions and achieve a natural and expressive appearance around the mouth area. Finally, the generated speech audio is combined with the modified video to produce a final video with synchronized lip movements and phonetic sounds.

Abstract: Aspects of the present disclosure include a method for verifying live user presence in an online session, comprising obtaining at least one video stream of a user captured during at least one previous presentation of at least one user interface (UI) on a display, obtaining input events representing previous user interactions with the UI, generating a behavioral digital fingerprint profile based on the video stream and the input events, initiating an online exam during the online session by providing, for presentation on the display, at least one additional UI, receiving at least one additional video stream of the user during the exam, receiving additional input events representing user interactions with the additional UI, determining exam-specific behavioral features based on the additional video stream and the additional input events, and verifying whether the user is taking the exam based on at least one of the features or the profile.

Abstract: Systems and methods provide a multidimensional data structure to tailor data storage and corresponding user output. Tailored graph data structures reduce the level of hallucination in composing an answer to a user using a large language model. In response to a user query of documents, the user query is translated to a vector. From the vector, the most relevant pieces of documents that are as similar as possible in semantics to the question of a person can be determined and communicated to the user. In one aspect, using the vector base (vector query applied to the documents), a graph can be generated in the form of nodes and connections between data in the documents.

Abstract: Disclosed herein are systems and method for a dynamic knowledge graph (KG) augmentation of an LLM. The method includes obtaining a LLM chat history between a user and the LLM enhanced with a KG database. The LLM chat history includes user queries, relevant knowledge data retrieved from the KG database, and LLM answers for each user query. The LLM answers are based at least in part on relevant knowledge data retrieved from the KG database. The method also includes applying a KG update MLM agent trained to: analyze the LLM chat history, KG database and the graph schema and to identify: (i) a missing knowledge data from the KG database, and/or (ii) a missing relationship between knowledge data nodes in the KG database. The method further includes after identifying missing knowledge data and/or the missing relationship, updating the KG database with new knowledge data and/or new relationship between nodes.

Abstract: Aspects of the present disclosure include a method for verifying live user presence in an online session. The method comprises providing, for presentation, a time-varying pattern and an instruction to a user to perform one or more requested user actions in synchrony with the pattern. The method further comprises receiving one or more data streams of the user during the presentation, extracting from the data streams feature information indicative of one or more detected user actions in the data streams, and generating one or more event signals based on the feature information. The method further comprises determining, based on the event signals and the pattern, a first measurement and a second measurement indicative of spatial correspondence and temporal alignment, respectively, between the detected user actions and the pattern, and verifying whether the user is a live person based on the first and second measurements.

Abstract: A system obtains, using a camera, a video stream of a user viewing a display. In response to changing visual characteristics of an object displayed on the display, the system detects visual changes on at least one surface of the user in the video stream. The system, based on a determination that the visual changes are within one or more thresholds, determines that the user is in front of the display. The system, based on a determination that the visual changes are not within the one or more thresholds, transmits a message that the user is not in front of the display.

Abstract: A system executes a first machine learning model to identify camera parameters of an input image. The system generates sets of shifted camera parameters by applying a set of predefined delta values to each of the camera parameters and calculates a plurality of homography matrices. The system executes a second machine learning model to identify a plurality of keypoints in the input image. For each respective set of the shifted camera parameters: the system projects the plurality of keypoints to planar coordinates using a homography matrix from the plurality of homography matrices that corresponds to the respective set of the shifted camera parameters, calculates a line of best fit that comprises the planar coordinates; and determines a total difference between the line of best fit and each of the planar coordinates. The system identifies a set of the shifted camera parameters with a minimal value of difference.

Abstract: Disclosed herein are systems and method for machine learning (ML)-assisted analysis of scientific or technical documents and associated software code, the method comprising: retrieving a scientific or technical document and a software code associated with the scientific or technical document, the software code including executable code and/or source code; analyzing the scientific or technical document using a trained paper analysis ML model configured to identify subject matter, logic, algorithms, and/or parameters of the scientific or technical document; analyzing the software code using a trained software analysis ML model configured to identify subject matter, logic, algorithms, and/or parameters of the software code; and comparing the algorithms and parameters of the scientific or technical document with the algorithms and parameters of the associated software code to identifying corresponding algorithms and parameters in the scientific or technical document and the software code.

Abstract: A system executes a first machine learning model to identify camera parameters of an input image, and calculates a homography matrix based on the camera parameters. The system determines, using the homography matrix, a first set of pixel coordinates of a plurality of keypoints on the input image. The system executes a second machine learning model of a second type to identify at least one of the plurality of keypoints in the input image. The system determines, from an output of the second machine learning model, a second set of pixel coordinates of the plurality of keypoints on the input image. The system calculates a difference between the first set of pixel coordinates and the second set of pixel coordinates. In response to determining that the difference is greater than a threshold difference, the system executes a third machine learning model to identify enhanced camera parameters of the input image.

Abstract: Aspects of the present disclosure include a method comprising receiving a video stream of a user during an online session, initiating a calibration by providing for presentation on a display a calibration element positioned at a first position on the display, detecting a gaze direction and/or head rotation of the user based on the video stream, determining a mapping between the first position and the gaze direction and/or head rotation, initiating a verification check during the session by providing for presentation on the display a visual stimulus element randomly positioned at a second position on the display, determining a change in the gaze direction and/or head rotation during the verification check, and verifying whether the user is in front of the display by determining whether the user reacted to the visual stimulus element based on the mapping, the change, and the second position. The verification check occurs after the calibration.

Abstract: Disclosed herein are systems and methods for training a text-to-speech machine learning model. A method includes: inputting training text into an acoustic model configured to generate an intermediate representation including predicted latent features for a vocoder model that further generates a waveform of speech reciting the training text; inputting a target waveform into a self-supervising learning (SSL) model configured to generate a vector representation of the target waveform, wherein the target waveform is true speech reciting the training text; extracting and summing SSL features from a plurality of layers of the SSL model; computing a loss between a sum of the SSL features from the SSL model and the predicted latent features from the acoustic model; updating, using backpropagation, weights of the acoustic model based on the loss; and executing the acoustic model with the updated weights on a test text to generate the intermediate representation.

Abstract: Disclosed herein are systems and method for a machine learning (ML)-based analysis of multiple simultaneous events in a video. In one aspect, a method includes: obtaining a video of objects that potentially involved in an activity; identifying the objects in the video by analyzing the video using a trained object detection ML model; cropping the video into a plurality of video clips; obtaining a list of roles for the activity and a list of actions associated the role; determining a role for each object involved in the activity by analyzing actions and/or interactions with the objects for each object by executing a trained role recognition ML model in each video clip; and evaluating a performance for each object in their role while performing the activity using a trained performance evaluation ML model.

Abstract: Disclosed herein are systems and method for monitoring race car pit crews and generating optimization recommendations. In one aspect, a method includes: receiving, from at least one camera, a video clip of a crew member performing service of a race car; detecting, using a machine learning algorithm, an error made by the crew member while performing the service; in response to determining that the error has been performed by the crew member more than a threshold number of times within a period of time, identifying a sequence of events leading up to the error; generating, using the machine learning algorithm, a recommended change in the sequence to prevent the error from reoccurring in a future service; and transmitting an instruction including the recommended change to a communicative device of the crew member.

Abstract: Disclosed herein are systems and methods for improving responses from LLMs by modifying chat history of at least one LLM. In one aspect, an exemplary method includes: obtaining a query; transmitting a prompt based on the query for input into a first LLM; obtaining a first response from the first LLM; displaying the query along with the first response from the first LLM; modifying a chat history of the first LLM based on obtaining a selection of at least one portion of the first response from the first LLM; combining the selected portions of the first response, the prompt, and the modified chat history into a new prompt for input into the second LLM; transmitting the new prompt for input into the second LLM; and obtaining and displaying a new response from the second LLM based on the new prompt.

Abstract: A system inputs a first audio sample into a co-speech engine that is configured to generate a first output data file comprising motion data of a virtual avatar over a period of time, wherein the motion data represents one or more gestures identified by the co-speech engine as corresponding to the first audio sample. The system extracts a first plurality of features from the first output data file. The system extracts a second plurality of features from a second output data file. The system determines a difference value by comparing the first plurality of features with the second plurality of features. The system updates weights associated with the co-speech engine based on the difference value between the first plurality of features and the second plurality of features. The system executes the co-speech engine with the updated weights on a third audio sample to generate a third output data file.

Abstract: Disclosed herein are systems and method for executing a text-to-speech machine learning model. A method includes: determining a first phoneme embedding from an input phoneme sequence; determining, using a text embedding model, a token-level embedding from an input word sequence, wherein the input phoneme sequence corresponds to the input word sequence; upsampling the token-level embedding into a second phoneme embedding; inputting both the first phoneme embedding and the second phoneme embedding in an encoder-decoder machine learning model configured to generate acoustic features for a vocoder model that produces a speech waveform; and executing the vocoder model to generate speech reciting the input word sequence.

Abstract: Disclosed herein are systems and method for ML-based analysis of racing communications. In one aspect, the method includes: obtaining a plurality of audio message between a plurality of race team members, converting the messages into text format, determining roles of speakers, including at least one of: determining roles of some speakers based on analysis of specific words and/or phrases, and determining roles of other speakers based on analysis of background noise patterns in audio messages, recognizing topics of messages by applying a third neural network trained on racing data, identifying a list of predefined keywordsin the text messages, determining a level of importance of each message based on the role of the speaker, the topic of the message, the predefined keywords, and a relationship of the message with other messages, and displaying the plurality of text messages based on the level of importance in a user interface.

Abstract: Systems and methods for detecting a malware injection interested processes. The method includes identifying one or more trusted processes, monitoring at least one thread associated with the trusted processes using at least one control point, detecting activity at the at least one thread based on the at least one control point and determining a timestamp of the detected activity, receiving from the trusted processes at least one execution stack corresponding to the timestamp and indicating the at least one control point used to monitor the at least one thread, applying a first malware detector to the at least one execution stack to generate a first verdict, collecting the first verdict and auxiliary information corresponding to the trusted processes at the given timestamp, and applying a second malware detector to the first verdict and the auxiliary information to generate a second verdict.