Selected technical writing on ML system design, retrieval infrastructure, and Bayesian modeling. Each piece focuses on implementation choices, trade-offs, and failure modes.
A system design walkthrough of BayesPilot as a production-style ML decision-support system, covering data contracts, training and calibration, deployment, and governance checks for reliable model updates.
A production-oriented retrieval architecture for RAG applications, focused on index design, latency-recall trade-offs, citation traceability, and operational patterns for evaluation and monitoring.
A research-focused analysis of SDM-CAR for spatial Bayesian modeling, showing how spectral parameterization improves robustness under misspecification, and how variational inference with MCMC supports uncertainty-aware prediction.