The rapid expansion of distributed and cloud-native systems has led to an exponential rise in operational logdatacharacterized by high volume, velocity, and variety. Conventional rule-based or signature-driven monitoring systems struggletogeneralize across diverse formats and dynamic environments, while deep-learning-based methods, though accurate, demandextensive computational resources and often behave as black boxes. This work introduces AI-Ops Log Anomaly Miner, a lightweight, parser-aware, and fully explainable framework for automatedloganalysis in resource-constrained environments. The system integrates log ingestion, normalization, and template miningwithhybrid feature extraction using TF-IDF and statistical window features. An unsupervised anomaly-detection layer—basedonIsolation Forest and ECOD algorithms—identifies deviations without labeled data, while an explainability modulegeneratesconcise human-readable reason codes. Implemented using Python, Streamlit, and SQLite, the prototype operates entirely offline and exports analytical summariesinCSV,JSON, and PDF formats. Benchmarking with open-source datasets such as HDFS and BGL demonstrates efficient detectionperformance under CPU-only conditions. The proposed system offers a transparent, deployable alternative to heavyweightAImodels, aligning with the operational goals of AIOps for dependable and interpretable anomaly detection. Keywords: AIOps, Log Analytics, Anomaly Detection, Unsupervised Learning, Isolation Forest, ECOD, TF-IDF, Drain3,Explainable AI, Parser-Aware Framework, Offline Deployment, Streamlit, SQLite