Abstract: A talking avatar generation system and method featuring one or more avatars, using natural language inputs received from the user to guide an artificial intelligence (AI) engine in generating talking avatar applications using natural language input from a user is disclosed. The method involves receiving a natural language input or prompt from the user which is then analyzed and parsed to convert it into a structured API request with commands and parameters. Using these commands, the application code is generated that integrates interactive avatars with the functionalities specified by the user. The avatar's personalities are created using prompt engineering techniques, enabling them to display distinct behaviors, emotions, and conversational traits. The generated code undergoes automated testing in a simulated environment to ensure reliability. Finally, the talking avatar application is deployed and displayed to the user, allowing real-time communication with the avatars.

Abstract: An avatar generation system and method generate dynamic and interactive three-dimensional avatars for educational interaction and a personalized virtual assistant. The avatar generation system utilizes a combination of 3D modeling, face reenactment, and text-to-speech module to produce 3D avatars with realistic movements and interactions. The avatar generation system utilizes prompts to guide an artificial intelligence (AI) engine in generating the avatars. The avatar generation system improves the scalability and personalization of avatars. Furthermore, the avatar generation system aims to provide a more efficient and quality-effective way for the generation of avatars. The use of algorithms for the automatic generation of movements and behaviors is introduced to produce more realistic and personalized animations.

Abstract: A content appropriateness validation system and method for validating content involves receiving content from a content management system 104 and conducting a comprehensive validation in accordance with predetermined metrics or other standards. In at least one embodiment, the standards are stored in a data storage and are inputs so that the programmatic control. The validation includes evaluating against standards such as grammar, coherence, factuality, engagement, age-appropriateness, and topic suitability, as well as identifying and filtering out inappropriate content related to sensitive topics like politics, sex, harassment, violence, hate, and self-harm. The system applies various algorithms and integrates a readability score generator to assess and assign a knowledge grade to the content. The validated content, along with the readability score, is then displayed to users on an online learning platform.

Abstract: A computer-implemented method is disclosed for transforming academic test performance into personalized feedback and learning recommendations. The method involves presenting an academic test to a user via a user interface of an online learning platform and receiving the user's submitted answers. The system accesses input parameters including historical user-performance data, correct answers, and coaching session data. The user's responses are compared with the correct answers to identify incorrect responses. A prompt generator creates a prompt to guide and constrain an AI engine in analyzing the test responses. The AI engine correlates the incorrect responses with historical performance data and coaching session information to detect learning patterns or recurring errors. Based on the identified patterns, the system generates personalized feedback and targeted learning recommendations to address specific learning gaps.

Abstract: This disclosure describes, in part, techniques for generating predictive models based on past transaction data and/or future-event data to predict: (i) when an account of user is expected to fall below a minimum balance associated with that user, and/or (ii) when an account has surplus funds that may safely be moved to a higher-yield account for some amount of time. In response to determining a predicted time at which an account is expected to fall below the minimum balance, the techniques may generate an offer to extend capital to the user prior to the predicted time. In response to determining that a user has or will have surplus funds, the techniques may generate an offer to move some or all of these surplus funds into a separate account providing a higher yield than the primary account of the user.

Abstract: In some examples, a predictive model is trained to determine account balances over which funds in user accounts are predicted to be surplus. Data of a first account of a user is input into the trained predictive model to determine a balance over which a portion of funds in the first account are predicted to be surplus funds for a period of time. A user device of the user presents a user interface that indicates predicted expenses and predicted income for a plurality of periods of time including the period of time. The user interface further indicates the surplus funds for the period of time and includes an interactive element for indicating disposition of the surplus funds. Based on receiving, via the user interface, an indication of an action to perform with respect to the surplus funds, at least one instruction is sent for performing the indicated action.

Abstract: In some examples, a predictive model is trained to determine account balances over which funds in user accounts are predicted to be surplus. Data of a first account of a user is input into the trained predictive model to determine a balance over which a portion of funds in the first account are predicted to be surplus funds for a period of time. A user device of the user presents a user interface that indicates predicted expenses and predicted income for a plurality of periods of time including the period of time. The user interface further indicates the surplus funds for the period of time and includes an interactive element for indicating disposition of the surplus funds. Based on receiving, via the user interface, an indication of an action to perform with respect to the surplus funds, at least one instruction is sent for performing the indicated action.

Abstract: In some examples, a system may receive transaction data indicating payments into a first account over time and payments out of the first account over time, the first account associated with a user. In addition, the system may access a data structure including historical account data that indicates variations in a balance of the first account over time. The system may determine an indicated minimum balance for the first account based on inputting the transaction data and the historical account data into a first predictive model configured to predict a minimum balance for enabling predicted payments out of a user account. The system may determine, based on a difference between a current balance of the first account and the predicted minimum balance for the first account, that the first account has a surplus of funds. Based at least on the surplus, the system may send an instruction.

Abstract: This disclosure describes, in part, techniques for automating transfers of funds from a deposit account of a user to a savings account of the user based on historical transaction data associated with the deposit account. Using historical deposits and withdrawals from the deposit account, a prediction may be made regarding deposits to, and withdrawals from, the deposit account that will occur in a future period. Based on that that information and a balance to maintain in the deposit account for the duration of the future period, the amount of funds to transfer from the deposit account to the savings account may be determined. The techniques may further include transferring the funds to the savings account once determined.

Abstract: In some examples, a system may receive transaction data indicating payments into a first account over time and payments out of the first account over time, the first account associated with a user. In addition, the system may access a data structure including historical account data that indicates variations in a balance of the first account over time. The system may determine an indicated minimum balance for the first account based on inputting the transaction data and the historical account data into a first predictive model configured to a predict a minimum balance for enabling predicted payments out of a user account. The system may determine, based on a difference between a current balance of the first account and the predicted minimum balance for the first account, that the first account has a surplus of funds. Based at least on the surplus, the system may send an instruction.

Abstract: This disclosure describes, in part, techniques for generating predictive models based on past transaction data and/or future-event data to predict: (i) when an account of user is expected to fall below a minimum balance associated with that user, and/or (ii) when an account has surplus funds that may safely be moved to a higher-yield account for some amount of time. In response to determining a predicted time at which an account is expected to fall below the minimum balance, the techniques may generate an offer to extend capital to the user prior to the predicted time. In response to determining that a user has or will have surplus funds, the techniques may generate an offer to move some or all of these surplus funds into a separate account providing a higher yield than the primary account of the user.

Abstract: This disclosure describes, in part, techniques for generating predictive models based on past transaction data and/or future-event data to predict: (i) when an account of user is expected to fall below a minimum balance associated with that user, and/or (ii) when an account has surplus funds that may safely be moved to a higher-yield account for some amount of time. In response to determining a predicted time at which an account is expected to fall below the minimum balance, the techniques may generate an offer to extend capital to the user prior to the predicted time. In response to determining that a user has or will have surplus funds, the techniques may generate an offer to move some or all of these surplus funds into a separate account providing a higher yield than the primary account of the user.

Abstract: This disclosure describes, in part, techniques for generating predictive models based on past transaction data and/or future-event data to predict: (i) when an account of user is expected to fall below a minimum balance associated with that user, and/or (ii) when an account has surplus funds that may safely be moved to a higher-yield account for some amount of time. In response to determining a predicted time at which an account is expected to fall below the minimum balance, the techniques may generate an offer to extend capital to the user prior to the predicted time. In response to determining that a user has or will have surplus funds, the techniques may generate an offer to move some or all of these surplus funds into a separate account providing a higher yield than the primary account of the user.