Forecasting and decision-support platforms
Building systems that move from raw operational data through Spark-class distributed processing, Iceberg/Nessie-style lakehouse architecture, normalized analytical layers, model execution, evaluation, APIs, and dashboards. Current work focuses on forecasting and decision-support platforms for demand planning, inventory health, pricing, and operational analytics, with reproducible pipelines and serving layers designed to make model outputs usable in production workflows.
AI-assisted development orchestration
Designing autonomous software development systems that route work across AI coding backends, preserve scope through structured task contracts, verify changes deterministically, and handle failure through explicit retry and escalation paths. A core architectural focus is the idea that dependable AI-assisted development emerges from contracts, verification, evidence, bounded execution, and governed progression — not prompt quality alone. Current work includes context contracts, backend capability harnesses, execution evidence systems, reconciliation workflows, and governed multi-backend orchestration across Codex CLI, Claude Code CLI, Gemini CLI, Aider, PI, and OpenCode.
Data platform modernization
Converting fragmented or legacy data workflows into scalable, maintainable platforms using cloud-native services, columnar data formats, reproducible pipelines, and production serving layers.
Document intelligence systems
Developing document intelligence systems that combine LLMs, embeddings, vector stores, and structured search to support code generation, research agents, operational assistants, and document-heavy business workflows.