Serverless GPU Architecture for Enterprise HR Analytics BDaaS Implementation

Guilin Zhang et al. publish a production-oriented BDaaS blueprint integrating a single-node serverless GPU runtime with TabNet for regulated enterprise HR analytics (accepted to IEEE BigData 2025).

• Main announcement: The paper describes a Helm-packaged, reproducible BDaaS blueprint that combines a single-node serverless GPU runtime with the TabNet interpretable model to support latency-sensitive, compliance-aware analytics in enterprise and government settings; benchmarks are reported on the HR, Adult, and BLS datasets.
• Details & context: The submission (arXiv v1: 22 Oct 2025; revised v2: 29 Jan 2026) presents benchmark results showing up to 4.5x throughput, 98x lower latency, and ~90% lower cost per 1K inferences versus Spark baselines, notes compliance overhead ~5.7 ms (p99 < 22 ms), and supplies a reproducible Helm package and decision framework; accepted to IEEE BigData 2025.