LuliDigital · Jan 2024
Lulidigital AI Operations Platform
AIAI Agent DevelopmentAI-Enhanced DesignAPI IntegrationArtificial IntelligenceAutomationAI Workflow DesignClaudeCanvaChatGPTContent MarketingCopywritingCRM
Currently an Extern @Pfizer·🟡 Open to consulting
Eager to Innovate with AI and Data Insights
I'm an AI and data enthusiast ready to elevate my career. Let's transform insights into action together!
Work samples
Dive into my portfolio showcasing projects focused on AI-powered insights and data extraction, designed to drive innovation and efficiency.
Lulidigital AI Operations Platform Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship 
Pfizer · ✅ Verified by Extern · ⏱️ In progress
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
AI & MLPythonDocument IntelligencePresentation Skills
About me
I'm an AI and data enthusiast ready to elevate my career. Let's transform insights into action together!
I began my career as a Clinical Nurse Anesthetist, where I developed strong skills in precision, collaboration, and problem-solving. My interest in technology led me to transition into AI and digital transformation. Today, I build AI-powered workflows, automation systems, and websites that help organizations improve efficiency. Recently, I completed Pfizer’s Advanced AI-Powered Document Insights & Data Extraction Externship, strengthening my experience in document intelligence, OCR, Retrieval-Augmented Generation (RAG), and AI-powered solutions.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
Pfizer
Experience
AI & Digital Transformation Consultant (Remote)
Freelance · Jan 2024 – Present
Clinical Nurse Anesthetist
Yekatit 12 Hospital Medical College · Sep 2016 – May 2023
Education
Addis Ababa University, Addis Ababa, Ethiopia
Bachelor's Degree in Clinical Nursing
Skills
AIData ExtractionDocument InsightsAnalytical ThinkingProblem SolvingArtificial IntelligenceAutomationBusiness IntelligenceCanvaChatGPTCollaborationCommunicationContent MarketingCopywritingCRMDashboardingDashboard DesignData AnalysisData VisualizationData Storytelling
Back to work
LuliDigital · Jan 2024
Lulidigital AI Operations Platform
AIAI Agent DevelopmentAI-Enhanced DesignAPI IntegrationArtificial IntelligenceAutomationAI Workflow DesignClaudeCanvaChatGPTContent MarketingCopywritingCRM
Overview
What I've accomplished
Designed and built Lulidigital, an AI-powered operations platform that combines workflow automation, website development, AI-assisted research, and digital marketing. Created practical systems that streamline business processes, improve content workflows, support client operations, and reduce repetitive manual tasks through AI-driven automation.
Outcome
✅ Verified by Extern · ⏱️ In progress
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
AI & MLPythonDocument IntelligencePresentation Skills
Overview
I developed an AI-powered document intelligence prototype with Pfizer, utilizing Optical Character Recognition (OCR), Large Language Models (LLMs), and Retrieval-Augmented Generation (RAG) to automate document workflows. This project showcased my ability to enhance enterprise PDF processing and solidified my skills in the AI domain.
What I've accomplished
I produced end-to-end artifacts: preprocessing experiments, extraction scripts and DataFrames, OCR engine comparisons with visualizations, and a LlamaIndex-based RAG pipeline that returned AI answers from ingested PDFs.
Project breakdown
I reviewed a mixed vendor file containing specifications, CoAs, regulatory and safety statements, technical sheets, tables and scanned pages. I identified varied date formats, mixed structured and unstructured data, and extraction/classification challenges for AI models.
In the image-enhancement task I processed a degraded scanned document in Colab. I applied multi-stage denoising (median blur, Non-Local Means, bilateral), CLAHE contrast on L channel, saturation restoration, luminance-only unsharp masking, EDSR 4x super-resolution, and Otsu thresholding, producing…
I extracted text and bounding boxes from an SDF PDF with PyMuPDF, used keyword matching and spatial grouping to link labels with values, and produced structured fields (manufacturing date, expiration date, lot number, product info).
I ran all three OCR engines on the same scanned pharmaceutical certificate, cleaned and aligned outputs, visualized detected text regions and bounding boxes, and produced a comparison report that found PaddleOCR best preserved layout and had the highest accuracy.
Google Docs
AccessThe project ingested a PDF, used sentence-based chunking and the all-MiniLM-L6-v2 embeddings, implemented hybrid retrieval (vector search plus BM25), and produced ranked passages and QC findings showing hybrid retrieval returned the most relevant, contextually accurate results.
Google Docs
AccessI tested three retrieval configurations on an SDF query, varied top_k and similarity threshold, and enabled an LLM reranker. Configuration C produced the clearest, most focused answer after filtering weaker matches and reranking retrieved chunks.
Google Docs
AccessA bundled pharmaceutical PDF was parsed page by page, pages were labeled for new-document boundaries and doc_type, ambiguous pages were reviewed and prompts refined, and a JSON metadata output recorded doc_id, page_in_doc, and doc_type for each page.
