Data Engineer and AI Engineer

Built an enterprise-grade real estate analytics Slack bot (Python, Flask, Slack Bolt, Snowflake, Redis, OpenAI GPT‑4o‑mini) that converts natural language requests into schema-validated SQL with multi-table join orchestration, semantic join-key recovery, and automated regeneration to eliminate invalid identifier errors. Engineered dynamic schema/profile caching with Redis, catalog-aware table selection, and background profiling pipelines to keep prompts and queries data-informed while maintaining cost controls. Delivered persistent, context-aware conversation workflows via the OpenAI Conversations API plus comprehensive diagnostics, security hardening, and proactive data-quality reporting. Built end-to-end machine learning pipeline for stock return prediction, processing 770K+ government contract records and 43K+ price data points from PostgreSQL; engineered financial and contract-based features with point-in-time data handling to prevent data leakage; developed and optimized XGBoost regression and classification models using grid search hyperparameter tuning; implemented modular Python framework enabling comparative analysis across financial-only, contract-only, and combined feature sets; created interactive visualization dashboard for exploratory analysis; documented methodology and results in research paper draft. Analyzed integrated US tax rate datasets to identify regional variances, uncovering discrepancies of over 2 percentage points in select counties, with concentrated misalignments in the Southeast and Southwest. Standardized and consolidated appraisal ratio datasets from government records into a clean, analysis-ready format supporting financial and policy insights. Identified usability, compliance, and technical gaps in a pharmaceutical document generator and proposed improvements in data validation, compliance handling, and AI-driven refinement to enhance accuracy, efficiency, and reliability. Built an automated data pipeline using web scrapers to collect, clean, and structure creator economy datasets, enabling accurate analysis and actionable insights.

Built an enterprise-grade real estate analytics Slack bot (Python, Flask, Slack Bolt, Snowflake, Redis, OpenAI GPT‑4o‑mini) that converts natural language requests into schema-validated SQL with multi-table join orchestration, semantic join-key recovery, and automated regeneration to eliminate invalid identifier errors. Engineered dynamic schema/profile caching with Redis, catalog-aware table selection, and background profiling pipelines to keep prompts and queries data-informed while maintaining cost controls. Delivered persistent, context-aware conversation workflows via the OpenAI Conversations API plus comprehensive diagnostics, security hardening, and proactive data-quality reporting. Built end-to-end machine learning pipeline for stock return prediction, processing 770K+ government contract records and 43K+ price data points from PostgreSQL; engineered financial and contract-based features with point-in-time data handling to prevent data leakage; developed and optimized XGBoost regression and classification models using grid search hyperparameter tuning; implemented modular Python framework enabling comparative analysis across financial-only, contract-only, and combined feature sets; created interactive visualization dashboard for exploratory analysis; documented methodology and results in research paper draft. Analyzed integrated US tax rate datasets to identify regional variances, uncovering discrepancies of over 2 percentage points in select counties, with concentrated misalignments in the Southeast and Southwest. Standardized and consolidated appraisal ratio datasets from government records into a clean, analysis-ready format supporting financial and policy insights. Identified usability, compliance, and technical gaps in a pharmaceutical document generator and proposed improvements in data validation, compliance handling, and AI-driven refinement to enhance accuracy, efficiency, and reliability. Built an automated data pipeline using web scrapers to collect, clean, and structure creator economy datasets, enabling accurate analysis and actionable insights.

Built an enterprise-grade real estate analytics Slack bot (Python, Flask, Slack Bolt, Snowflake, Redis, OpenAI GPT‑4o‑mini) that converts natural language requests into schema-validated SQL with multi-table join orchestration, semantic join-key recovery, and automated regeneration to eliminate invalid identifier errors. Engineered dynamic schema/profile caching with Redis, catalog-aware table selection, and background profiling pipelines to keep prompts and queries data-informed while maintaining cost controls. Delivered persistent, context-aware conversation workflows via the OpenAI Conversations API plus comprehensive diagnostics, security hardening, and proactive data-quality reporting. Built end-to-end machine learning pipeline for stock return prediction, processing 770K+ government contract records and 43K+ price data points from PostgreSQL; engineered financial and contract-based features with point-in-time data handling to prevent data leakage; developed and optimized XGBoost regression and classification models using grid search hyperparameter tuning; implemented modular Python framework enabling comparative analysis across financial-only, contract-only, and combined feature sets; created interactive visualization dashboard for exploratory analysis; documented methodology and results in research paper draft. Analyzed integrated US tax rate datasets to identify regional variances, uncovering discrepancies of over 2 percentage points in select counties, with concentrated misalignments in the Southeast and Southwest. Standardized and consolidated appraisal ratio datasets from government records into a clean, analysis-ready format supporting financial and policy insights. Identified usability, compliance, and technical gaps in a pharmaceutical document generator and proposed improvements in data validation, compliance handling, and AI-driven refinement to enhance accuracy, efficiency, and reliability. Built an automated data pipeline using web scrapers to collect, clean, and structure creator economy datasets, enabling accurate analysis and actionable insights.

Built an enterprise-grade real estate analytics Slack bot (Python, Flask, Slack Bolt, Snowflake, Redis, OpenAI GPT‑4o‑mini) that converts natural language requests into schema-validated SQL with multi-table join orchestration, semantic join-key recovery, and automated regeneration to eliminate invalid identifier errors. Engineered dynamic schema/profile caching with Redis, catalog-aware table selection, and background profiling pipelines to keep prompts and queries data-informed while maintaining cost controls. Delivered persistent, context-aware conversation workflows via the OpenAI Conversations API plus comprehensive diagnostics, security hardening, and proactive data-quality reporting. Built end-to-end machine learning pipeline for stock return prediction, processing 770K+ government contract records and 43K+ price data points from PostgreSQL; engineered financial and contract-based features with point-in-time data handling to prevent data leakage; developed and optimized XGBoost regression and classification models using grid search hyperparameter tuning; implemented modular Python framework enabling comparative analysis across financial-only, contract-only, and combined feature sets; created interactive visualization dashboard for exploratory analysis; documented methodology and results in research paper draft. Analyzed integrated US tax rate datasets to identify regional variances, uncovering discrepancies of over 2 percentage points in select counties, with concentrated misalignments in the Southeast and Southwest. Standardized and consolidated appraisal ratio datasets from government records into a clean, analysis-ready format supporting financial and policy insights. Identified usability, compliance, and technical gaps in a pharmaceutical document generator and proposed improvements in data validation, compliance handling, and AI-driven refinement to enhance accuracy, efficiency, and reliability. Built an automated data pipeline using web scrapers to collect, clean, and structure creator economy datasets, enabling accurate analysis and actionable insights.