Modern serverless workflows deployed across the edge-cloud continuum are facing placement challenges where functions with varying data and computational requirements are often placed far from required data or on insufficiently capable hardware. The stateless execution model forces functions to exchange data through high-latency storage service APIs rather than direct communication, creating performance bottlenecks where data movement dominates execution time in data-intensive domains such as in smart cities, healthcare, machine learning, and earth observation. Current container orchestration platforms like Kubernetes and serverless platforms like Knative mainly consider CPU and memory resources while neglecting data locality, network topology, and specialized hardware capabilities. Current scheduling approaches in this direction are limited to static weight configurations that cannot adapt to changing workload patterns, manual data dependency annotations forcing developers to predict access patterns at development time, and simplified network models ignoring actual topology relationships. We develop a data-locality-aware scheduler extending Kubernetes with a Multi-Criteria Decision-Making (MCDM) framework that integrates with Knative for serverless workloads and that dynamically weighs workload characteristics, cluster topology, bandwidth constraints, data placement, and node capabilities while adapting MCDM priority weights in real-time based on function requirements. Our system allows for dynamic per- invocation data dependency specification through CloudEvents and implements automatic storage discovery with topology-aware network modeling. Experimental evaluation against the default Kuberentes scheduler demonstrates an improvement from 25.1% to 73.3% in local data access, a 46.7-79.4% reduction in network transfers, and a 87.2% reduction in cross-region transfers, with minimal scheduling overhead averaging 3.1 seconds across our tested workflows.