Industrial automation increasingly relies on multi-agent AI, yet evaluation remains difficult due to task complexity and data confidentiality. We present AssetOpsBench-Live, a demo of a competition-ready platform for real-time, privacy-preserving evaluation of multi-agent AI in industrial contexts. The platform integrates AssetOpsBench, which measures six dimensions of multi-agent performance and performs automated failure-mode discovery, with Codabench, which supports reproducible, code-oriented competitions. End users first validate agents locally, then submit containerized code for execution on hidden industrial scenarios. Instead of raw trajectories, the system provides quantitative scores and clustered failure modes (e.g., reasoning--action mismatch, step repetition), enabling participants to identify failures, apply targeted improvements, and iteratively resubmit. By combining competition-based engagement with actionable diagnostics, AssetOpsBench-Live delivers reproducible, real-time insights reflecting real-world industrial constraints.

Maintenance of mission-critical industrial assets is frequently hindered by fragmented data, inconsistent record-keeping, and limited access to analytical expertise, resulting in reactive rather than predictive practices. We present \textit{CodeReAct}, an AI-powered agentic framework deployed in large-scale facilities to automate event analysis and work order (WO) management.CodeReAct extends the ReAct paradigm by embedding executable Python code within the Thought--Action--Observation (TAO) loop, enabling natural language interaction, grounding heterogeneous alerts and work orders into structured Business Objects (BOs), and dynamically invoking analytic functions for forecasting, anomaly correlation, and maintenance recommendations. This architecture reduces manual data science intervention, improves adaptability, and supports reuse across asset types. Deployed in a mission-critical data center and productionized in Maximo, CodeReAct manages pumps, chillers, AHUs, compressors, cooling towers, and other mechanical and electrical systems. Evaluation with 36 representative maintenance utterances showed that outer-loop reflection and adaptive temperature improved task completion by up to 20%, while ablation studies confirmed the importance of reasoning in addition to code execution. Business validation revealed seasonal failure patterns, bundling opportunities, and predictive accuracy trends. In production, site engineers reported 25--40% faster diagnostics, fewer unplanned downtime events, and reduced reliance on specialized analysts. Lessons learned highlight the importance of structured BOs for grounding analytics, runtime safeguards to mitigate hallucinations, and adaptive model control for consistent execution. These results demonstrate how deployed agentic AI can deliver measurable business value in predictive and strategic maintenance planning.

Operation recommendations and automation for various industries rely on optimization over the top of the real-time Internet of Things data streams. Such recommendation applications have become imperative for industrial manufacturing, such as petroleum, chemicals, and food processing. We build a novel system of user-in-the-loop multiobjective optimization under the initial uncertainty of the optimization objectives, wherein the uncertainty is iteratively resolved via user feedback. We propose an interactive optimization system wherein both business and operational goals become defined as the optimization processes and where objective selection is incorporated as part of the optimization procedure. We show that such a solution exists during the iteration process if the feasibility space is not empty initially and is constrained by industry operations. Using an oil sands application, we demonstrate this approach and compare modeling results in values, response to business and operational priorities, and interpretability to the weighted sum optimization.

Data quality assessment is an essential process of any data analysis process including machine learning. The process is time-consuming as it involves multiple independent data quality checks that are performed iteratively at scale on evolving data resulting from exploratory data analysis (EDA). Existing solutions that provide computational optimizations for data quality assessment often separate the data structure from its data quality which then requires efforts from users to explicitly maintain state-like information. They demand a certain level of distributed system knowledge to ensure high-level pipeline optimizations from data analysts who should instead be focusing on analyzing the data. We, therefore, propose data-quality-aware dataframes, a data quality management system embedded as part of a data analyst's familiar data structure, such as a Python dataframe. The framework automatically detects changes in datasets' metadata and exploits the context of each of the quality checks to provide efficient data quality assessment on ever-changing data. We demonstrate in our experiment that our approach can reduce the overall data quality evaluation runtime by 40-80% in both local and distributed setups with less than 10% increase in memory usage.

Time series value forecasting using machine learning models utilizing time series features has recently got good attention of Time series analytics community. This paper proposes an automated feature learning mechanisms to filter out most useful features from hundreds of available features for time series prediction problems. The paper further proposes a novel mechanism to dynamically filter features that are most suitable for the given input time series data. With such mechanisms we create pipeline consisting of most useful features for given input data and increases the performance of the prediction model. Our proposed mechanism first, groups well known features for time series analysis, generates and assigns the features importance score using multiple scoring configurations. Once scores are assigned, features are filtered using a threshold that is derived using reference feature score and Critical Difference diagram. The filtered features are subsequently analyzed based on the characteristics of the input dataset. We show using experimental results that our approach of input data based dynamic feature selection improves the overall performance of machine learning models compared to the case where dynamic feature extraction is not applied prior to modeling.

The success of IoT depends on our ability to solve challenging problems which were previously infeasible. One of the most critical challenges in IoT space is the preventive maintenance in industrial manufacturing processes to maximize equipment availability and durability. Traditionally, preventive management only follows less cost-effective strategies, say time or usage based management. With large amount sensor data harvested from IoT, we can develop much more intelligent predictive maintenance based on accurate machinery failure prediction. In this paper, we develop a framework named DoM (Doctor for Machines) to produce the best predictive model for several oil and gas industry engagements. Our framework is built in the form of pipeline that allows us to generate multiple models simultaneously with parallel computing. We configure the modeling process by assigned different machine learning tasks as such sampling, feature extraction, modeling, and post-processing into the pipeline. The pipeline forms a machine learning graph workflow. With the automation of execution all the tasks in the workflow, the client can easily choose a best predictive model fitting their failure tolerance. We have successfully applied DoM to six data sets, and identified valuables insights on the best practices of creating predictive model. Our clients have successfully achieved financial benefits by applying DoM to develop predictive maintenance schemes to their expensive equipment.

Data Quality (DQ) has been one of the key focuses as Data Analytics and Artificial Intelligence (AI) fields continue to grow. Yet, data quality analysis has mostly been a disjointed, ad-hoc, and cumbersome process in the overall data analysis workflow. There have been ongoing attempts to formalize this process, but the solutions that have come out are not universally applicable. Most of the proposed solutions try to address the problem of data quality from a limited perspective and suc-cessfully address only a subset of all challenges. These solutions fail to translate to other domains due to a lack of structure. In this paper, we present DQLearn, a toolkit for structured data quality learning. We start by presenting the core principle on which we build our library and introduce the four components that provide a solid base to address the needs of the data quality problem. Then, we showcase our automation structure - "Workflows", and the two optimization techniques equipped with it, that help the users to structure their learning problem very easily. Next, we discuss four important scenarios of the DQ Workflows in the overall life-cycle. Finally, we demonstrate the utility of the proposed toolkit with public datasets and show benchmark results from optimization experiments.

This paper proposes a verifiable imputation process and an enabling tool for univariate time series. Common ad-hoc and case-specific imputation are not enough to ensure high quality and effective imputation. We adopt the similar verification logic of supervised learning. We use artificial missing sampling as the test set to estimate a set of imputers' performances and use the estimated performances to select the best imputer. To ensure the correctness of selection, we analyze the impact of various factors on estimation accuracy. Those factors are missing rate, size of artificial missing data and patterns, selected imputers, and noise level. We propose a two-step verifiable imputation process to integrate all of the steps. With this process, we can always leverage the most suitable imputer to achieve a high quality of imputation without tedious and error-prone data cleaning efforts. We implement the tool as a Python package, with many imputers with their unique capabilities and a API. We automate the imputation through a standard process, which returns imputed results and detailed rationales of selection along with quality metrics.

For several years now, scientists have been proposing numerous models for defining anything "as a service (aaS)", including discussions of products, processes, data & information management, and security as a service. In this paper, based on a thorough literature survey, we investigate the vast stream of the state of the art in Everything as a Service (XaaS). We then use this investigation to explore an integrated view of XaaS that will help propose approaches for migrating applications to the cloud and exposing them as services.

With the wide availability of products and services through popular e-commerce platforms and dozens of similar offerings to choose from, there is a need to accurately assess and evaluate the quality of offerings. Several studies have shown that consumer feedback is an important source of information. This paper presents: (a) consumer Rating as a Service (RaaS) -- a building block service that can be used to add the consumer feedback lifecycle feature in the development of e-commerce platforms, (b) an approach to evaluate the quality of composite offerings based on the aggregation of consumer ratings using the composition structure and component relationships. Benefits of the proposed service include reduced development effort, shorter delivery time and a fine-grained aggregation of consumer ratings for composite offerings even with limited ratings.

With the increasing of the popularity of Service oriented Software Development, we have identified there is a need to systemically reduce the complexity and increase the robustness of developed Service systems through a guided development process. We choose under design (UD) and over design (OD) as core value assets to uniformly represent the target of managing the human errors and deficiencies during a development process. In the study of a Web Service architectural selection case, based on the ideology of Value Driven Design, we proposed the Problem-Value-Constraint (PVC) approach as the framework of minimizing under design and over design covering both IT concerns including functionalities and qualities and Business concerns including value, cost, price and usage. Our PVC solution connects the characteristics including knowledge management, information transformation and business value. Partially through service design patterns, our study showed that PVC is able to outline both business and technical characteristics at the same time keep the simplicity of model structures.

Service brokering has an increasingly prominent role in bridging the gap between business requirement and technology enablement. We propose the concept of service value brokering (SVB) to fulfil the possible missing linkages between business and technology layer. In this paper, we modeled a SVB Web service as an integration of two layers with the business interface (BIF) and the technical interface (TIF). With this distinction, Web service compositions can map to two layers of compositions at both BIF and TIF levels. We notice that any partial consideration on the consistency of either BIF or TIF layers would likely leave mismatching situations on the other layer. We employ SVB to solve these mismatching situations. With the help of SVB, we address the needs of coherent business planning and IT implementation in a model drivenmanner. Finally, we illustrate the feasibility of our approach in the development of a modern cloud-based tourism e-commerce platform.

With the growing interest in cloud computing, more and more businesses are looking not only to migrate their applications to the cloud but also to transform them to better leverage capabilities that are provided by cloud platforms or to enable new business models that are facilitated by the cloud. One problem clients face in this area is a lack of experience and knowledge as to how best to accomplish this transformation. We propose a Cloud Transformation Advisor (CTA) which helps users to select appropriate enablement patterns from a knowledge base of best practices when performing transformation planning. This knowledge base uses a structured representation to capture application information, cloud platform capability information, and enablement pattern information in order to facilitate pattern selection. We describe this representation and a mathematical model which leverages it to choose the "best" combination of patterns for a given transformation problem. We present an example which illustrates the approach, and describe the usage of the CTA.

Software architectures evolve over time, and so do the models that represent them. For a domain like Service Oriented Architecture (SOA) this is particularly true because most SOA solution designs are based on modification of existing assets that change over time. However, today there exists only limited work that reasons about this evolution. In this work we present our framework for traceability of evolving architectures that we apply for SOA solution design. Our design approach is based on an iterative process that utilizes a set of solution patterns to guide architects in the SOA solution design. Our approach utilizes historical data about pattern enablement and uses that to guide architects in selecting the right patterns. To ensure that the right patterns are used, we use a template matching approach that enforces conformance by allowing only the right set of artifacts to be composed together. We demonstrate how our framework can be applied to compose and trace evolving SOA solutions based on three views - the artifact view, profile view and compliance view.

This paper provides an artifact-pattern-matching framework and mathematical model to analyze the dynamic behaviors of the SOA solution design in model driven fashion and provide recommendations for optimal solution pattern enablement for solution artifacts. The artifact-pattern-matching system can be dynamically tuned based on the practitionerspsila final selections of there commendations. Specifically, we propose a set of solution patterns to guide SOA solution architects through the process of consuming and configuring SOA artifacts for composing SOA solutions. The resulting multi-dimensional cascading flagging method is also presented in this paper. As an example, impact analysis patterns are used as solution patterns to support traceability enablement. We present some future directions of leveraging reinforcement learning algorithms to enrich the design quality analytics of SOA solution.

In this paper, we present an architecture-centric assessment approach for model evaluation over reference architecture to quantitatively estimate architecture maturity and quality. Such assessment is essential to support design-level refinement for an enterprise solution. To achieve this analytic goal, we select a nine-layer SOA solution stack (S3) as reference architecture, and introduce the necessary mathematical definitions and formulation. The baseline for such assessment is a model template composed of S3 solution patterns. A template is the starting point of creating a design model. The selection of such template will largely determine the architecture properties of the final SOA solution.The maturity analysis is carried out at different granularity levels (architecture building block, architecture layer, and architecture model) to justify the 'completeness' of a design. The quality assessment is accomplished through a set of quality-indicators to justify the 'goodness' of an architecture based on the relationships of architecture building block instances. Finally, using UML 2.0 to capture the model of S3, we provide a real assessment prototype developed over IBM RSA platform.

This paper considers basic bounds on the overhead of link-state protocols in mobile ad hoc networks. Hierarchical protocols are known for their good scalability properties, and hence this paper considers a two-level hierarchical protocol. In such protocols, nodes need to keep track of shortest path information, link states and cluster membership. Two types of overheads are considered; the memory needed to store routing-related information, including link-states and cluster membership, and the control messages that need to be exchanged to keep track of the changes in the network. Memory overhead is important practically for dimensioning network nodes, while message routing overhead is important since it reduces the effective capacity of the network to carry user data (vis-a-vis control data). The scalability properties of the message routing overhead are analyzed for different modes of network scaling. Practical implications, such as optimal cluster size, average/fixed memory requirement and routing protocol parameter selections are discussed.

The rapid adoption of model-driven design (MDD) methodology in SOA-based solution design requires an adaptive tooling environment that can systematically improve designers' productivity. Ideally, the environment should be flexible enough to both handle frequently changing requirements and support new features without intensive coding efforts. In this paper, we provide a coding-free enablement framework to realize such extensible tooling environments based on a mathematical abstraction of key models in SOA solution design using graph theory definition. This abstraction formalizes the SOA modeling logic and semantics, and also guides the implementation of an extensible and customizable tooling environment. As a case study, we illustrate how our framework is able to transform the development style from Java programming to text editing through our implementation of a UML 2.0 based SOA modeling environment using IBMpsilas Rational Software Architect (RSA) development platform.

In this paper, information theoretic techniques are used to derive analytic expressions for the minimum expected length of control messages exchanged by proactive routing in a two-level hierarchical ad hoc network. Several entropy measures are introduced and used to bound the memory size necessary for the storage of the routing tables. The entropy rates of the topology sequences are used to bound the communication routing overhead-both the interior routing overhead within a cluster and the exterior routing overhead across clusters. A scalability analysis of the routing overheads with regard to the number of nodes and the cluster size is provided under three different network scaling modes. Finally, practical design issues are studied by providing the optimal cluster sizes that asymptotically minimize (i) the memory requirement for each cluster head; (ii) the total control message routing overhead.

This paper presents a mathematical and simulative framework for quantifying the overhead of reactive routing protocols, such as dynamic source routing and ad hoc on-demand distance vector, in wireless variable topology (ad hoc) networks. A model of the routing-layer traffic, in terms of the statistical description of the distance between a source and a destination, is presented. The model is used to study the effect of the traffic on the routing overhead. Two network models are analyzed; a Manhattan grid model for the case of regular node placement, and a Poisson model for the case of random node placement. We focus on situations where the nodes are stationary but unreliable. For each network model, expressions of various components of the routing overhead are derived as a function of the traffic pattern. Results are compared against ns-2 simulations, which corroborate the essential characteristics of the analytical results. One of the key insights that can be drawn from the mathematical results of this paper is that it is possible to design infinitely scalable reactive routing protocols for variable topology networks by judicious engineering of the traffic patterns to satisfy the conditions presented in this paper.

This paper presents a new mathematical and simulative framework for quantifying the overhead of a broad class of reactive routing protocols, such as DSR and AODV, in wireless variable topology (ad-hoc) networks. We focus on situations where the nodes are stationary but unreliable, as is common in the case of sensor networks. We explicitly model the application-level traffic in terms of the statistical description of the number of hops between a source and a destination. The sensor network is modelled by an unreliable regular Manhattan (i.e. degree four) grid, and expressions for various components of the routing overhead are derived. Results are compared against ns-2 simulations for regular and random topologies, which corroborate the essential characteristics of the analytical results. One of the key insights that can be drawn from the mathematical results of this paper is that it is possible to design infinitely scalable reactive routing protocols for variable topology networks by judicious engineering of the traffic patterns to satisfy the conditions presented in this paper.