Prudential Financial

Task-oriented evaluation of knowledge graph (KG) quality increasingly asks whether an ontology-based representation can answer the competency questions that users actually care about, in a manner that is reproducible, explainable, and traceable to evidence. This paper adopts that perspective and focuses on gap and overlap analysis for policy-like documents (e.g., insurance contracts), where given a scenario, which documents support it (overlap) and which do not (gap), with defensible justifications. The resulting gap/overlap determinations are typically driven by genuine differences in coverage and restrictions rather than missing data, making the task a direct test of KG task readiness rather than a test of missing facts or query expressiveness. We present an executable and auditable benchmark that aligns natural-language contract text with a formal ontology and evidence-linked ground truth, enabling systematic comparison of methods. The benchmark includes: (i) ten simplified yet diverse life-insurance contracts reviewed by a domain expert, (ii) a domain ontology (TBox) with an instantiated knowledge base (ABox) populated from contract facts, and (iii) 58 structured scenarios paired with SPARQL queries with contract-level outcomes and clause-level excerpts that justify each label. Using this resource, we compare a text-only LLM baseline that infers outcomes directly from contract text against an ontology-driven pipeline that answers the same scenarios over the instantiated KG, demonstrating that explicit modeling improves consistency and diagnosis for gap/overlap analyses. Although demonstrated for gap and overlap analysis, the benchmark is intended as a reusable template for evaluating KG quality and supporting downstream work such as ontology learning, KG population, and evidence-grounded question answering.

We present ACORN-Edu, a deadline-aware prefetching system combining content-defined chunking (CDC), delta synchronization, and multi-factor scheduling for learners with intermittent connectivity. The scheduler prioritizes content using deadline urgency, chunk reuse, file size, and connectivity-aware finishability scoring, with weights tunable per deployment scenario. Evaluation across two scenarios, nightly Wi-Fi (~900,000 KB budget) and spotty cellular (intermittent bursts, ~24,000 KB budget), shows ACORN-Edu achieves 90% hit-rate vs. 60% baseline (Wi-Fi) and 20% vs. 0% (cellular) with minimal bandwidth overhead (4% and 1%). Statistical tests confirm highly significant improvements (p < 0.001, Cohen's d > 3.0). Ablation studies reveal urgency and reuse drive hit-rate gains under favorable connectivity, while all components optimize efficiency under scarcity. Full artifacts available at https://github.com/divineiloh/ACORN-Edu.

Hybrid data products combine the content, applicative, and consumption components of data sharing in a singular consumable package that any data consumer can leverage with little or no effort. Architectures facilitating hybrid data products, like data mesh and data fabric, address the increasing complexity of data integration and sharing across silos. At the same time, demand for agentic AI solutions—those that act on behalf of the user—is on the rise. Hybrid data products have particular relevance for group insurance and retirement solution platforms, given the availability of predictive and treatment effect models, semantic simulated events for scenario testing, and personalized decision recommendations. Extra care should be taken to ensure that data products in the financial domain do not perpetuate model or sampling bias and that they adhere to industry regulations, from data privacy—where applicable—to risk provisioning. Although clearly Patterned for Financial Services, these concerns are secondary to the stability, accessibility, and usability of hybrid data products at scale.

Insurance risk prediction is very difficult, due to the diversity and richness of customer information. The traditional statistical or independent deep learning models often face the challenge of trading off accuracy, interpretability, and deployment efficiency. We introduce a Hybrid DNN–DevOps Framework, as it integrates use of CNN, GRU and Attention to more efficiently capture the spatial-temporal relationships in claim data and manage for ongoing optimization using DevOps automation. With advanced features like containerized deployment, ML flow tracking, and Prometheus-based monitoring we can offer real-time reliability. Extensive experiments demonstrate the efficiency and effectiveness of proposed model with 98.4% accuracy, 98.15% F1-score for speech VR, and only 48 ms latency comparing with state-of-art methods such as Random Forest(94.8%) and CNN–LSTM (96.8%). The results verify the scalability of the framework, its stability to data drift and the ability to run efficiently in dynamic cloud environments. The combination of automated retraining pipelines, and predictive intelligence, enhance claim analysis, fraud discovery and policy optimisation making the proposed system competitive for advanced insurance analytics. The findings provide a solid basis for intelligent, explainable and scalable risk prediction systems in banking and insurance sectors.

In smart grid operations, Neighborhood Area Networks (NANs) play a crucial role in ensuring low latency communication but face challenges like congestion and cyberattacks. This paper introduces a multi-controller Software-Defined Networking (SDN) framework integrating a proactive prediction model using Graph Neural Networks (GNN) and a reactive Deep Q-Network (DQN) for real-time routing optimization. The framework features a hybrid failover mechanism for seamless transitions across various technologies (Wi-SUN, LoRa, ZigBee, 5G, and wired PLC) and includes an Intrusion Detection System (IDS) for secure path selection. Implementation using Mininet, NS-3, and real-world datasets demonstrates improved failover latency, throughput, packet delivery ratio, and attack detection accuracy compared to existing methods.