Data Engineer

This project demonstrates a production-style data engineering pipeline built using AWS, Snowflake, and dbt to process Airbnb data from raw ingestion to analytics-ready datasets.The pipeline is designed with scalability and maintainability in mind, leveraging a medallion architecture (Bronze - Silver - Gold) to progressively refine data for business use. Tech Stack Cloud Storage - AWS S3 Data Warehouse - Snowflake Transformation - dbt Language - Python 3.12, SQL Version Control - Git Key Features • Incremental Processing for scalable data pipelines • SCD Type 2 Snapshots for historical tracking • Modular dbt Models for maintainability • Reusable Macros for dynamic transformations • Data Quality Tests for reliability • Analytics-Ready Tables for reporting What This Project Demonstrates • End-to-end data pipeline design • Medallion architecture implementation • Incremental processing at scale • Historical tracking with SCD Type 2 • Production-style dbt structure • Strong data modeling practices More details can be found in the repo below https://github.com/Opsy-Daizy/dbt_snowflake_aws_project_end_to_end

This project evaluates how effectively a Voice AI channel supports digital service access, especially for users who may face accessibility barriers such as rural location, disability, or first-time digital use. The analysis uses a DuckDB SQL pipeline with a dbt-style layered model. Raw CSV files are loaded as source views, cleaned into staging views, aggregated into intermediate session-level metrics, and combined into an analysis-ready fact table: fact_voice_ai_sessions. The final model links voice session metadata, user attributes, AI performance metrics, turn-level error signals, and application outcomes. This enables evaluation of completion rates, drop-offs, AI friction, escalation behavior, and accessibility gaps across user segments and service channels. The project's main analytical goal is to identify where Voice AI improves access, where it underperforms compared with Web and USSD, and how high-severity AI degradation can be reduced by 40% while improving application completion. What the pipeline enables * Session-grain fact table with one row per voice session. * Integrates users, voice sessions, turns, AI metrics, and application outcomes. * Measures completion, drop-off, escalation, error severity, and accessibility segments. * Supports dashboard views for Voice vs Web/USSD comparisons and friction analysis. * Defines a practical 40% error-reduction target for high-severity Voice AI degradation. Technology: DuckDB SQL models with dbt-style source, staging, intermediate, and mart layers.