---
type: work
area: work
status: active
date: 2026-05-03
created: 2026-05-03
updated: 2025-10-23
tags:
  - work
---
### Project: Automated Power BI Data Export & Processing Pipeline

Developed a high-performance, resilient data pipeline in Python to automate the extraction, processing, and validation of large-scale datasets from Power BI. The system is engineered for both ad-hoc and scheduled batch operations, significantly improving data accessibility and reliability for downstream analytics.

- Automated Data Extraction: Architected a robust system to execute dynamic DAX queries against Power BI Analysis Services, handling data retrieval for custom time periods.
- High-Performance Architecture: Implemented a dual-mode orchestration engine, featuring a background processing manager with a priority queue to handle large, multi-month exports efficiently without blocking main execution threads. This design optimizes memory usage by offloading large data frames to temporary Parquet files.
- Resilient & Fault-Tolerant: Engineered an enhanced retry mechanism that distinguishes between different failure stages (export vs. validation), enabling smart retries and ensuring high operational uptime. Run state is persisted to allow seamless recovery from interruptions.
- Data Integrity & Validation: Integrated an automated validation module to perform integrity checks on exported data, generating summary reports and snapshots that are appended to a consolidated Excel workbook for easy auditing.
- Modern Tooling: Leveraged Python for core logic, the .NET ADOMDClient for native Power BI connectivity, and Apache Parquet for efficient data storage. The application is designed as a portable command-line tool using PyInstaller.

### STALGEN-AI: Automated Test Scenario Generation for Automotive Software
#### Project Description:

Spearheaded the development of STALGEN-AI, a Python-based application that automates the generation of software test scenarios from functional requirement documents for the automotive industry. This tool significantly accelerates the validation process for complex systems like Stellantis CRONOS/OMS, ensuring compliance with ISO 26262 standards.

#### Key Contributions & Features:

- Full-Stack Pipeline Development: Engineered an end-to-end pipeline that parses Word documents (.docx) into Markdown, extracts critical requirements using regex, generates structured test cases in JSON format via Large Language Models (LLMs), and populates them into a pre-defined Excel template.
- Advanced LLM Orchestration: Implemented a sophisticated prompt engineering system utilizing Chain-of-Thought (CoT) reasoning and few-shot examples to steer LLMs (via Ollama) toward generating accurate and structured test scenarios. The system includes JSON validation and graceful fallback mechanisms to handle generation failures.
- Intuitive User Interface: Developed a clean, user-friendly web interface using Streamlit that allows users to upload documents, select from available LLM models, and download the final test suite, making the tool accessible to non-technical users.
- Robust & Modular Architecture: Designed a modular Python application with clear separation of concerns for parsing, data extraction, LLM interaction, and file conversion, ensuring maintainability and scalability.
#### Technologies Used: Python, Streamlit, Ollama, Pandoc, OpenPyXL, Pydantic.