Stock Time-Series Orchestrator
This private working project is a local Python platform for stock time-series research. It uses an LLM coordinator for planning and explanation while keeping diagnostics, forecasting, backtesting, and scoring in deterministic Python modules.
System idea
The architecture separates orchestration from analytics. Agentic AI helps coordinate the workflow, but the actual data-science steps remain testable and reproducible:
- universe selection
- data-quality checks
- spectral diagnostics
- SARIMAX and Fourier-enhanced forecasting comparisons
- walk-forward backtesting
- model agreement and uncertainty review
- opportunity scoring
Why it matters
This project reflects a practical agentic AI pattern: use an AI coordinator to manage analytical work, but keep the numerical methods auditable. That boundary is important for financial analytics, where explanations should not replace reproducible evidence.
Portfolio relevance
The project shows agentic workflow design, time-series forecasting, analytics engineering, and a clear separation between AI-generated reasoning and deterministic model evaluation.