Volunteer computing systems exploiting large amounts of geographically dispersed resources on the Internet for solving complex scientific problems. However, scheduling scientific workflows in a fully decentralised way and low overhead is a challenging task in these environments. To counter this challenge, this paper presents a fully decentralised proximity-aware workflow-scheduling policy for these environments. The proposed scheduling consists of three phases. In the first phase, each workflow application is partitioned into sub-workflows in order to minimise data dependencies among them. The second phase of the workflow-scheduling algorithm finds some resources to execute each sub-workflow. These resources are selected based on Quality of Service (QoS) constraints of the workflow, load balancing and proximity of resources. Each workflow can have QoS constraints in terms of minimum CPU speed and minimum RAM or hard disk requirements. In the third phase, sub-workflows will be executed on each resource based on local scheduling algorithm to minimise the partial makespan. The proposed scheduling policy focuses on the reduction of communication overhead to improve the performance of I/O-intensive and data-intensive workflows. Simulation results show that the proposed workflow-scheduling policy improves the average response time of scientific workflows up to 53.6% under a moderate load.

Automated Machine Learning (AutoML) reduces the technical requirements for applying machine learning, yet existing systems still demand technical expertise, optimize inefficiently, and rarely support conversational use. We present a conversational AutoML system with a five-agent architecture coordinated via natural language, tuning Random Forest and gradient-boosting models (XGBoost, LightGBM, and CatBoost) with Monte Carlo Tree Search (MCTS) and Bayesian optimization under repeated cross-validation conditions. Its distinctive component is an ensemble coordinator: specialist optimizer agents tune model families concurrently, and a coordinator fuses their outputs with the strongest baseline into a stacking ensemble, deploying it only when it improves on the best single model by cross-validation. This separates two measurable benefits: a speed-up from parallel search and an accuracy gain from ensemble fusion. On six benchmark datasets (five classification and one regression), the multi-agent coordinator is best or tied on every dataset and the only configuration with a positive average gain over defaults (+0.20 pp, versus −0.69 pp for single-agent MCTS); its search is $2.3\times$ faster than the single-agent Bayesian baseline in aggregate, and it is systematically better than single-agent MCTS across datasets (Wilcoxon signed-rank, p = 0.031). While individual dataset gains are observed, none reach statistical significance when adjusted for multiple comparisons. For classification tasks, we additionally report F1-score, precision, recall, and AUC-ROC (Area Under the Receiver Operating Characteristic Curve).