Middle Python Developer (IRC303938)

GlobalLogic is working on an advisory engagement to design an AI-powered search assistant for project documentation accumulated over roughly 20 years, currently stored on an isolated, internet-disconnected file server. The dataset spans 2-3 TB including test artifacts, with core documents in the 1-10 GB range and about half the volume in the legacy .xls format. The goal is to let developers and SMEs query specification documents in natural language and get answers assembled across multiple related customization projects, instead of manually tracing code history and cross-referencing documents by hand. The target architecture connects the document store (likely SharePoint) to a generative AI search layer, with NotebookLM as the initial reference point and enterprise alternatives such as Vertex AI Agent Builder under evaluation.

Job Description

• 3+ years of commercial experience with Python, including work on data processing and ETL pipelines
• Experience parsing and transforming document formats (Excel including legacy .xls, PowerPoint, Word, PDF) using libraries such as pandas, openpyxl, xlrd, python-docx, python-pptx
• Experience with OCR technologies for extracting text from images and embedded objects (e.g. Tesseract, Google Vision API, Azure Computer Vision) 
• Understanding of RAG (Retrieval-Augmented Generation) architectures, embeddings, and vector search
• Experience integrating with cloud storage and collaboration platforms via API (SharePoint / Microsoft Graph API or similar)
• Written and spoken English proficiency equivalent of B2 or higher
• Ability to work with incomplete requirements during an advisory and PoC phase
• Self-reliant, comfortable proposing and testing technical solutions with limited direction

Nice to have:

• Familiarity with NotebookLM, LangChain, LlamaIndex, or Vertex AI Agent Builder
• Experience designing incremental indexing or delta-update mechanisms for search systems
• Experience working with enterprise clients in isolated or air-gapped network environments
   

Job Responsibilities

• Design and implement a format-conversion pipeline for legacy files (.xls, .ppt) into AI-readable formats (.xlsx, PDF, or Markdown)
• Build extraction logic for unstructured data inside Excel files, including text in shapes via drawing-layer analysis and OCR processing of embedded images
• Develop and validate an incremental update mechanism so only new or changed files are reprocessed
• Implement data cleansing logic to exclude obsolete or duplicate files (marked "old", "BKUP") from search results
• Evaluate NotebookLM against alternative platforms (e.g. Vertex AI Agent Builder) and document findings to support the Advisory Phase comparison
• Build PoC search functionality supporting natural language queries across multiple customization specification documents