
Giovanni De Marco developed prompt classification and KPI-driven analytics features for the belgio99/smartfactory repository over a two-month period. He built an end-to-end Jupyter notebook workflow that classifies user prompts using a Gemini 1.5 API, generating labeled datasets and integrating a query generator to automate KPI engine and predictor API calls. His work included robust error handling, ambiguous KPI resolution, and post-processing checks to ensure reliable analytics and reporting. Leveraging Python and data engineering skills, Giovanni refactored project structure, improved code documentation, and addressed stability issues, resulting in a scalable, well-organized backend that supports advanced time series analysis and reporting.

Month: 2024-12 — Delivered KPI-driven improvements, robust query generation, and reliability enhancements in belgio99/smartfactory. Key features shipped include Query Generation Improvements with ambiguous KPI handling, KPI Engine integration with predictor and post-processing checks, and report generation support. Major bug fixes addressed stability issues in KPI retrieval, parsing, class consistency, and prompts. These efforts yield stronger business value through more accurate KPI-based queries, reliable time-window predictions, and improved readiness for analytics and reporting.
Month: 2024-12 — Delivered KPI-driven improvements, robust query generation, and reliability enhancements in belgio99/smartfactory. Key features shipped include Query Generation Improvements with ambiguous KPI handling, KPI Engine integration with predictor and post-processing checks, and report generation support. Major bug fixes addressed stability issues in KPI retrieval, parsing, class consistency, and prompts. These efforts yield stronger business value through more accurate KPI-based queries, reliable time-window predictions, and improved readiness for analytics and reporting.
Month 2024-11 Key Deliverable: Prompt Classification and KPI Engine Integration for RAG in belgio99/smartfactory. Implemented an end-to-end notebook workflow that classifies user prompts into labels (predictions, new_kpi, report, kb_q, dashboard, kpi_calc) using a Gemini 1.5 few-shot API. Generated and tested labeled datasets, introduced a query generator to build KPI engine/predictor API calls, and refined validation datasets and notebook examples. Reorganized notebooks under a classifier subdirectory to support RAG enhancements and future KPI integration, establishing a scalable foundation for KPI-driven analytics.
Month 2024-11 Key Deliverable: Prompt Classification and KPI Engine Integration for RAG in belgio99/smartfactory. Implemented an end-to-end notebook workflow that classifies user prompts into labels (predictions, new_kpi, report, kb_q, dashboard, kpi_calc) using a Gemini 1.5 few-shot API. Generated and tested labeled datasets, introduced a query generator to build KPI engine/predictor API calls, and refined validation datasets and notebook examples. Reorganized notebooks under a classifier subdirectory to support RAG enhancements and future KPI integration, establishing a scalable foundation for KPI-driven analytics.
Overview of all repositories you've contributed to across your timeline