Digital Commons Network^™

Llm-Driven Autonomous Agent For Comprehensive Automated User Experience Testing, Johannes Start, John Lunney

Defensive Publications Series

Current front-end testing methods are slow, expensive, brittle, and limited to specific predefined scenarios without considering high-level tasks and user intent. This disclosure describes a large language model (LLM) powered autonomous agent designed for comprehensively testing the User Experience (UX) of a software solution by taking into account high-level tasks and user intent. The LLM used to power the agent can be trained or fine-tuned on datasets of UI state representations paired with corresponding interaction steps, task instructions paired with successful interaction sequences, examples of efficient and inefficient interaction patterns, and descriptions of principles for good UI design. The agent …

Optimizing User Experience By Continual Automated Assessment Of Usability Metrics Obtained Using A Large Language Model, Johannes Start, John Lunney

Defensive Publications Series

Current software testing methods and metrics do not provide a comprehensive, objective, and automated assessment of the UX in real time, thus making it challenging to be proactive about identifying and addressing usability issues that have a negative impact on the UX and business goals. This disclosure describes techniques that leverage the capabilities of large language models (LLMs) to provide a deep, comprehensive, and nuanced understanding of the UX of a software solution that takes into account the user’s intent and context. Preprocessed data about interaction analytics and contextual information obtained with user permission can be analyzed via an LLM …

Dlmp: A Deterministic Lightweight Memory Protector For Rowhammer-Resilient Computing Platforms, Mostefa Sayeh Mr

Defensive Publications Series

This document introduces DLMP (Deterministic Lightweight Memory Protector), a standalone memory protection architecture designed to counter RowHammer vulnerabilities in DRAM-based systems. DLMP enforces deterministic, low-overhead protection by tracking row access activity and enforcing per-row decay-based thresholds using integer-only logic and an FSM-based control f low. It is platform-agnostic and ideal for use in embedded systems, NPUs, and other latency sensitive computing environments. DLMP ensures that no row can be activated beyond a safe threshold by dynamically assigning row “temperature” states (Hot, Warm, Cool) and triggering enforcement through the Row Cooldown Enforcer (RCE). This document is …

Document-Level Understanding For Browser Contexts, Hp Inc

Defensive Publications Series

This article presents a browser-native assistant for advanced contextual understanding of web content. The system features real-time DOM analysis, semantic saliency scoring, and structured memory, enabling users to query and summarize web pages in natural language. By leveraging real-time browser events, the assistant dynamically adjusts context relevance to improve the accuracy and relevance of its responses.

Field Prism Height Calibration, Hp Inc

Defensive Publications Series

This disclosure relates to the field of robotic layout systems, particularly those utilizing Robotic Total Stations (RTS) for position tracking and projection.

Printer Orientation Detection Via Existing Shaid Sensors, Hp Inc

Defensive Publications Series

Printer orientation detection via existing shaid sensors is a solution that make use of existing ink tank shaid pins in the printer to double up as printer orientation detection, for detecting possible tilts in the printer that may cause various issues.

Partial Encapsulation Of Thermal Interface Material To Improve Performance, Anonymous

Defensive Publications Series

This disclosure presents a novel Thermal Interface Material (TIM) encapsulation method designed to enhance thermal contact between electrical or optical components and associated cooling structures such as heatsinks, test slugs, or housings for optical modules. While existing high-performance TIMs offer superior thermal conductivity, their adoption has been constrained due to fragility, particle detachment, or lack of adhesive properties, resulting in compromised reliability and performance. The disclosed solution employs partial encapsulation of the TIM between thin protective foils, such as aluminum, copper, or polymer films. This encapsulation secures the TIM, prevents shifting, contains particle debris, and provides mechanical robustness. Additionally, strategically …

Automating Geospatial Analysis Using Large Language Models And Code Execution In A Persistent Environment, Patrik Blohmé, Ananya Gupta, Gearoid Murphy, Sameera Ponda, Danusch Merrikh-Yazdi, Gia Jung, Renee Johnston, Marcin Sieniek

Defensive Publications Series

Sustainability professionals, urban planners, clean energy developers, and other professionals encounter substantial challenges in efficiently analyzing and utilizing geospatial data for decision-making related to location opportunities, accessibility, and asset placement. Traditional approaches involve complex, multi-step workflows that require specialized expertise and time-consuming analysis across disparate data sources, resulting in substantial delays before actionable insights can be derived. This disclosure describes techniques for automating geospatial workflows over multiple datasets from varied sources, facilitated by large language model (LLM) agents. The techniques enable users to efficiently query geospatial data and receive insights through an AI-driven process that minimizes complexity and reduces time …

Real-Time Conversational Diagram Generation System, Gregoire Magendie, Erwan Zerhouni, Stephane Martin, Toby Godwin

Defensive Publications Series

A system is proposed herein that provides for the ability to automate the real-time conversion of spoken ideas into visual diagrams, thereby dynamically creating and updating visuals, such as flowcharts, organizational charts, or the like during online collaboration sessions, such as online meetings or the like. The proposed system integrates an Automatic Speech Recognition (ASR) Engine with Large Language Model (LLM) -based semantic parsing, live rendering, and persistent memory to ensure multi-modal collaboration and continuity across collaboration sessions. By reducing the cognitive load on participants and eliminating manual diagram management, the proposed system accelerates the design and documentation of complex …

Fictitious Data-Based Watermarking System

Defensive Publications Series

Systems and methods are provided for watermarking via unique synthetic datasets. A candidate model may be trained with a poison dataset including synthetic information. The synthetic information may relate to a fictitious place, person, or scenario, for example. A set of one or more inference questions associated with the poison dataset may be determined. The candidate model may be tested based on the set of one or more inference questions. A confidence may also be generated. The confidence may be associated with one or more identification metrics associated with the candidate model.

Method, System, And Computer Program Product For Large Language Model (Llm)-Enabled Searching, Jia Yi Kuo, Jianyang Huang, Karthik Gopalakrishna, Bhuvanjeet Singh Gandhi, Paromita Guha, Ashutosh Kumar Chandel, Arpita Agrawal, Ritu Saxena

Defensive Publications Series

A method includes: chunking each resource into chunks; inputting the chunks into a model; for each chunk, generating an embedding vector; receiving a query message from a user device; inputting the query message into the model; generating a query embedding vector for the query message; executing a semantic search on the query embedding vector by: inputting the query embedding vector and the embedding vector for each chunk to the model; for each chunk, generating a chunk similarity score; determining a subset of chunks based on the chunk similarity score for each chunk of the resource; inputting the subset of chunks …

Comprehensive Analysis And System Design For Earth's Rotational Kinetic Energy Conversion, Michael @ Gaiacortex

Defensive Publications Series

This paper presents a detailed analysis and full-system design for a novel method of continuous electric power generation by harnessing Earth's rotational kinetic energy through its interaction with the planet’s own non-rotating magnetic field. Challenging a longstanding assumption in electrodynamics that Lorentz-force-induced electromotive forces cancel in Earth-bound reference frames, this work demonstrates that specific conductor topologies, magnetic permeability configurations, and low magnetic Reynolds number conditions enable persistent emf generation. The system is stationary, emission-free, and designed for long-term deployment with minimal environmental footprint. A proof-of-concept prototype yields measurable microvolt-level output, validating the theoretical framework. The design is modular and scalable, …

Unified Container Platform For Data Management Across The Entire Lifecycle, Puneet Mahajan, Suryakant Gokhale, Race Wright, Romit Kusarye, Ashley Brown, Vijay Jain

Defensive Publications Series

Managing the lifecycle of complex data processing operations involves significant challenges in ensuring consistency, handling dependencies, optimizing resource usage, and maintaining data quality across large-scale data assets that are not met by existing solutions. This disclosure describes a cohesive approach to data management by encapsulating related data components into unified containers with automated versioning, access control, and dependency tracking, and operation monitoring for greater efficiency, consistency, and data discoverability across the entire lifecycle of the data. A unified container can serve as a simplified grouping mechanism that enables key tasks to be managed at the container level instead of being …

Automating Tasks In Video Games Via Intelligent Agent, Ilia Labzovsky, Danny Karmon

Defensive Publications Series

This document describes techniques that enable a computing device (e.g., a personal computer, laptop, smartphone, smart watch, server system, gaming console, tablet, wearable device, virtual reality (VR) headset, augmented reality (AR) glasses, etc.) to automate repetitive tasks in video games using an intelligent software agent (e.g., artificial intelligence (AI), machine learning model, etc.). Video games often necessitate repetitive, time-consuming tasks for player progression, leading to user fatigue and diminished enjoyment (e.g., loot farming, resource gathering, experience point (XP) grinding, repetitive combat encounters). In some instances, the player may have to complete the same tedious tasks multiple times to reach engaging …

Custom Thin-Coat Designs For Reflective Waveguides In Augmented Reality Wearables, Jun Yang, Alexander Koshelev, Christophe Peroz

Defensive Publications Series

The use of reflective waveguides in augmented reality (AR) displays can reduce see-through transmission, such that a person facing the user can perceive an uneven brightness of the user’s face near the eye region. Uneven face brightness can cause some social and cosmetic discomfort. This disclosure describes a family of graded thin-film coats for reflective waveguides with multiple SKUs, each tailored to serve users of specific skin tone, color, and/or texture. The coating for the eye-pupil expander (EPE) and output coupler (OC) regions of the lens are tuned to achieve certain reflectance and wavelength-dependence criteria that render the prisms on …

Context-Aware Content Recommendation During Travel, Shatakshi Gupta, Sayan Acharya, Ajay Karthik Nama Nagaraj, Netri, Saumya Rangani, Parantap Sharma

Defensive Publications Series

A context-aware content recommendation process for an application running on a computing device (e.g., a smart television, a mobile computing device) may provide context-aware content recommendations to a user on the go (e.g., during travel). The context-aware recommendation process may determine that a user on the go may have different content consumption patterns and computing device access. In some implementations, the computing device may be shared with other users (e.g., a smart television in a hotel room, a tablet belonging to a friend). The context-aware content recommendation process may interface with device sensors included on the computing device that may …

Semantic Policy Summaries And Semantic Explanations For Explainable Ai In Networking, Dennis Lanov

Defensive Publications Series

Proposed herein is an explainability technique, referred to herein as a "Semantic Policy Summaries and Semantic Explanations" technique, for artificial intelligence (AI) -driven networking that translates complex model decisions into clear, human-readable explanations grounded in real-world network concepts. Instead of abstract outputs like feature scores, the proposed technique delivers goal-oriented justifications for AI network-related decisions/actions. The proposed technique integrates directly into a control loop, supports policy-based triggering, and logs explanations with decision metadata for audit and regulatory review. Further, the proposed technique aligns with requirements from the European Union (EU) AI Act and the proposed United States (U.S.) SAFE Innovation …

Gain Movie Insights Using Generative Artificial Intelligence, Akash Verma, Vineet Kumar, Abhay Kumar Gupta, Vivek Yadav, Szu-An Lu, Meng-Ting Tsai, Ya Hsien Liu, Chu-Feng Lien

Defensive Publications Series

A television (TV) application may present various types of media content of interest to a user. The media content may have different formats such as streaming video and audio. The types of media content may include, but are not limited to, movies, television shows, live sporting events, news items, short form videos, and music. In addition, or in the alternative, a variety of media content providers may deliver various types of media content for viewing by the user. The TV application may deliver a customized viewing experience to a user that spans the diverse types of media content provided by …

"G" Shaped Non-Functional Pad To Reduce Common Mode Noise, Jianquan Lou, Alpesh U Bhobe, Haiwen Lu

Defensive Publications Series

Provided herein are techniques for designing "G" shaped non-functional pads in vias of SERDES (Serializer-Deserializer) channels to reduce common mode noise that can be caused by twice the Nyquist frequence of SERDES signals, which commonly causes SERDES radio frequency issues for 25.78 Gigahertz (GHz) and 26.56GHz signals. Advantageously, the "G" shaped non-functional pad design proposed herein can reduce common mode noise while not impacting SERDES differential signals.

Agency Protocol: A Decentralized Trust-Building System Using Domain-Specific Merit And Economic Stakes To Incentivize Promise-Keeping In Agent-To-Agent Interactions, David Joseph

Defensive Publications Series

The Agency Protocol defines a decentralized trust infrastructure in which autonomous agents cooperate by issuing cryptographically‑signed, domain‑scoped promises and verifiable assessments. Each promise carries a dual‑component stake—an at‑risk credit deposit and a non‑refundable operational‑cost fee—that is locked in content‑addressed storage and returned or slashed according to assessment consensus. Assessments are merit‑weighted; an agent’s merit is a non‑transferable, domain‑specific reputation score updated by matrix‑factorized aggregation of historical outcomes, while credits form a fully transferable internal currency. Merit discounts future stake requirements, creating a positive feedback loop that makes honest behaviour utility‑maximising.

External facts enter the system through a decentralised oracle …

Monolithic Assembly For Augmented Reality Eyewear With Improved Comfort, Daniel Effinger, Matthew Valente

Defensive Publications Series

This disclosure describes techniques to integrate multiple distinct modules of a display system for augmented reality eyewear into a single, monolithic unit or carrier. Per the techniques, a highly integrated, compact assembly replaces conventional multi-module systems. This is achieved through the design and manufacturing of a waveguide (WG) carrier that serves as a monolithic housing. The carrier is precisely molded to incorporate physical characteristics and features necessary for passively installing various optical and electronic components. Such components may include the waveguide, lenses for a light engine assembly (LEA), and lenses for a camera module. Active alignment (AA) of critical sensors …

Multi-Agent Architecture With Central Intelligent Agent, Shashwat Anand, Rohan Chaudhury

Defensive Publications Series

This disclosure describes a multi-agent architecture featuring a centralized intelligent agent (orchestrator agent) for complex enterprise environments, e.g., partner interfaces for a cloud computing solutions provider. While agent-driven interfaces are feasible, there are challenges in agent discovery and coordination, leading to fragmented solutions and slower pace of development. By providing a central agent that coordinates with various domain-specific functional agents, the techniques of this disclosure provide a cohesive mechanism for managing and orchestrating agents. The agent is built to service various user personas and uses this context to provide high-fidelity assistance. Agents are built with awareness of access control lists …

Location-Specific Artificial Intelligence Models, Ajay Prasad

Defensive Publications Series

Virtual assistants powered by large language models (LLMs) are capable of interacting with users via chat, voice, or other interfaces and can answer arbitrary queries from the user. However, LLMs may not have access to updated location-specific information. This disclosure describes techniques for integrating location-specific AI models into virtual assistants. This allows virtual assistants to provide hyperlocal information and personalized responses by switching between general-purpose and specialized models with user permission. Location-specific models are trained with data relevant to particular places or businesses. The described techniques enhance user experience by providing accurate and up to date location-specific information appropriate to …

Scope 4 Impact And Reporting Methodology, Marisol Palmero, Snezana Mitrovic, Suresh Krishnan

Defensive Publications Series

The concept of Scope 4 emissions, though not officially recognized in the Greenhouse Gas Protocol, presents a strategic opportunity to enhance sustainable operations. By focusing on "avoided emissions", based on good or best practices, organizations can influence and reduce Scope 2, and even Scope 3 emissions across their value chain. Techniques are proposed herein that seek to reduce energy consumption from networking devices. The proposed techniques could be extended to other practices within organizations. By reducing energy consumption, there is a direct link to improve energy efficiency, optimize resource usage, and ultimately contribute to an organization's overall sustainability goals. The …

Secure Coding Practices: An Auditor's Checklist For Enterprise Applications For Faang By Ramkumar Bharathan, Ramkumar Bharathan

Defensive Publications Series

This report serves as a comprehensive guide to critical secure coding practices, presented as an actionable checklist for developers and security teams operating within large-scale enterprise environments. It reflects the rigorous security standards and pragmatic approaches typically observed within leading technology organizations. The core objective is to foster a "shift-left" security paradigm, integrating robust security measures throughout the Software Development Lifecycle (SDLC) to proactively identify and mitigate risks rather than reacting to vulnerabilities post-deployment. The checklist systematically addresses key areas, including input validation, authentication, access control, cryptographic practices, error handling, logging, secure file management, and system configuration. For each practice, …

A Hybrid Ai-Powered Voice-Enabled Medical Practice Assistant With Dynamic Triage Routing And Multi-Modal Patient Interaction, Patrick Proctor

Defensive Publications Series

We present a novel implementation of a voice-enabled medical practice virtual assistant that combines real-time speech processing, intelligent medical triage routing, and multi-modal patient interaction capabilities. Our system employs a hybrid AI architecture that dynamically selects between different language models based on query complexity classification, reducing operational costs while maintaining high-quality responses for complex medical inquiries. The system incorporates a deterministic medical triage decision tree based on established nursing protocols, enabling automated routing of urgent cases while maintaining HIPAA compliance. Additionally, we introduce an innovative multi-step interaction pattern for location-aware services, seamlessly transitioning between voice input, location detection, and turn-by-turn …

Low-Bandwidth, High-Sensitivity Telemetry Signal To Optimize Osc Launch Power During Deployment And Regular Operation, Anonymous

Defensive Publications Series

The present disclosure relates to methods and systems employing a low-bandwidth, high-sensitivity telemetry signal integrated into optical line systems to dynamically optimize the launch power of an Optical Supervisory Channel (OSC) during both deployment and normal operation. Traditionally, OSC power management faces challenges such as receiver overload on short spans and insufficient received power on longer spans, requiring manual provisioning. The disclosed telemetry signal independently and reliably measures optical span loss across an extended range (up to approximately 50 dB), significantly exceeding the measurement capabilities of conventional OSC signals. Using telemetry-derived span-loss data, OSC launch power can be automatically and …

Natural Language Based Intuitive Business Intelligence Tool For End Users, Nilesh Patel, Naresh Jasotani

Defensive Publications Series

This disclosure describes a natural language processing (NLP)-based business intelligence (BI) tool designed to help business users analyze structured data without requiring technical expertise. The tool helps overcome challenges faced by business users, such as reliance on technical teams for data access, the complexity of traditional BI tools, and the need for technical skills to interact with databases. The NLP-based tool performs automated relationship discovery for entities within a data store, enables automated prompt and query generation, provides business-oriented data views in an intuitive user interface for data exploration, question asking, and report generation. Implementation of the tool removes the …

A System For Predictive Crowd Management Using Coupled Analysis Of Crowd State And Environmental Information, Mithun Kumar S R

Defensive Publications Series

Some crowd management systems analyze crowd dynamics, such as density and flow, in isolation from the physical environment. The disclosed system can couple a Dynamic Crowd State Vector, which may quantify real-time crowd metrics, with a Hyper-Local Condition Vector, which may quantify environmental factors such as surface integrity, passage width, and lighting. A machine learning model can process the real-time interaction between these two vectors to generate a unified, predictive risk score for a specific geographic zone. The resulting coupled analysis may facilitate the proactive identification of context-specific hazards, such as moderate crowds in some conditions, which may aid in …

Color-Coded Lane Guidance System For Dynamic Traffic Flow Optimization, Jacques Mouton

Defensive Publications Series

Traffic flow can be sub-optimal when drivers lack real-time, lane-specific information. This disclosure describes a system for dynamic lane guidance, for example, within a navigation application on a computing device (e.g., a smartphone, smart watch, wearable device, augmented reality glasses, an in-vehicle infotainment system, etc.). In some implementations, the system can utilize a probabilistic inference engine, such as a Bayesian inference engine, to analyze multiple data sources. These sources may include real-time traffic data, historical data, predictive models, and crowdsourced information. This analysis can be used to calculate a probability of congestion, delay, and/or accident risk for individual lanes. These …