Currently an Extern @Pfizer
As a senior in Computer Science, I specialise in Artificial Intelligence and sharpen my skills whenever I have the opportunity to create innovative solutions that push boundaries.
Dive into my portfolio showcasing projects in Artificial Intelligence and database architecture, reflecting my skills and creativity.
Breaking Games · Jul 2026 · ✅ Verified by Extern
Turn six messy CSVs into an analytics database and a Q4 dashboard that drives real business decisions. SQL + Claude.
Pfizer · ✅ Verified by Extern · ⏱️ In progress
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
As a senior in Computer Science, I specialise in Artificial Intelligence and sharpen my skills whenever I have the opportunity to create innovative solutions that push boundaries.
I am a senior studying Computer Science with a focus on Artificial Intelligence. Currently, I am gaining hands-on experience as an extern at Breaking Games, where I work with SQL and database architecture.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
Pfizer
Experience
Founder
Association Neb lã Bũmbu · February 2023 – Present
Marine Conservation Externship
National Geographic Society + TNC · September 2025 - November 2025
Consumer Insights Data Analytics Externship
Beats by Dre · June 2024 - July 2024
Education
Ashesi University Berekuso, Eastern Region
BSc. Computer Science · Class of 2027
Skills
✅ Verified by Extern
Turn six messy CSVs into an analytics database and a Q4 dashboard that drives real business decisions. SQL + Claude.
I transformed six disorganized CSV files into a structured analytics database and developed a Q4 dashboard to inform strategic business decisions. This project involved utilizing SQL and collaborating with the team at Breaking Games to enhance data-driven insights.
I produced a joined analytics database, a denormalized dashboard_metrics table, a static Chart.js HTML dashboard, and a Holiday Game Plan containing cleaned revenue figures, adjusted cart-abandonment rates, channel conversion comparisons, and three Q4 recommendations.
✅ Verified by Extern · ⏱️ In progress
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
The externship explored AI-driven document intelligence for enterprise PDF workflows, combining OCR, large language models, and retrieval-augmented generation. The work involved researching model architectures, document-processing pipelines, and techniques for extracting structured data from unstructured documents. Deliverables included a technical synthesis and recommendations for building a
.jpg)
I produced a catalog of ten pharmaceutical document types, enumerated six-plus date formats, and documented specific extraction challenges such as multiple date meanings, historical dates in body text, boilerplate noise, redactions, and near-identical layouts that impeded type classification.
I examined a multi-document vendor PDF, identified ten document types and six-plus date formats, noted multiple dates with different meanings and a misleading "supersedes" reference, and listed AI extraction challenges found in the file.
The module work involved loading pharma-style datasets in Colab, cleaning and standardizing columns with Pandas, normalizing nested JSON into a table, and applying image preprocessing (grayscale, CLAHE, median blur, Otsu) to improve OCR input. Deliverables included Colab notebooks and processed…
The project processed multi-page SDF PDFs, extracted text and bounding boxes with PyMuPDF, used regex and anchor phrases to find manufacturing, expiration, and document dates plus lot numbers and vendor name, and verified matches visually.
The project evaluated three OCR engines on the same scanned supplier PDF. I ran each engine, compared field-level recognition for lot numbers, dates, and vendor text, analyzed output structure and bounding-box data, and measured processing times. The report found all three matched key fields,…
I processed a three-page PDF, chose overlapping chunking, generated embeddings with sentence-transformers/all-MiniLM-L6-v2, implemented vector retrieval, and produced exact answers to questions about test methods and storage conditions from the document.
