Pfizer logoPfizer Externship via Extern
Barisua Ntor-ue portrait

Aspiring Data Architect

I turn messy data into meaningful insights, driving business decisions through SQL and database architecture.

Work samples

Explore how I transformed messy data into a powerful analytics database and dashboard during my externship at Breaking Games.

  • The Unofficial University Guide: Production RAG Pipeline
    The Unofficial University Guide: Production RAG Pipeline

    Engineered an end-to-end RAG pipeline to process a multi-document corpus of unstructured housing regulations and student data Implemented semantic vector search using all-MiniLM-L6-v2 embeddings with strict distance thresholds (<0.65) to enforce source attribution and eliminate LLM hallucinations

    PythonChromaDBGroq
  • Agentic Music Recommendation System
    Agentic Music Recommendation System

    Designed a 4-part agentic architecture (retriever, agent, evaluator, recommender) processing natural-language queries against a 10,000+ item catalog via REST Built reliability guardrails with structured logging and 42 unit tests, reducing hallucinated outputs by 91% with a 100% pass rate

    PythonREST

About me

I turn messy data into meaningful insights, driving business decisions through SQL and database architecture.

I'm on the cusp of launching my career, currently gaining valuable experience through an externship in SQL and database architecture at Breaking Games. I thrive on transforming data into actionable insights.

Externships

Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship

In progress

Experience

AI Training Engineer

AfterQuery · Apr 2026 6 Jun 2026

AI Trainer

Handshake · Dec 2025 6 Present

Data Analytics Extern

Breaking Games · Jul 2026 6 Present

Education

University of St. ThomasSt. Paul, MN

Bachelor of Arts in Computer Science · Class of 2027

Skills

SQLDatabase ArchitectureData Analytics

The Unofficial University Guide: Production RAG Pipeline

Engineered an end-to-end RAG pipeline to process a multi-document corpus of unstructured housing regulations and student data Implemented semantic vector search using all-MiniLM-L6-v2 embeddings with strict distance thresholds (<0.65) to enforce source attribution and eliminate LLM hallucinations

PythonChromaDBGroq
View all work

Agentic Music Recommendation System

Designed a 4-part agentic architecture (retriever, agent, evaluator, recommender) processing natural-language queries against a 10,000+ item catalog via REST Built reliability guardrails with structured logging and 42 unit tests, reducing hallucinated outputs by 91% with a 100% pass rate

PythonREST
View all work

Full Stack Web Application

Modeled a 5-table normalized SQL schema using SQLAlchemy ORM, cutting query times by 55% vs. flat-file architectures Authored 7+ RESTful Python/Flask API endpoints integrated with a React frontend

ReactPythonFlaskSQLAlchemy
View all work

✅ 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

The project built a prototype pipeline for extracting structured data from enterprise PDFs using OCR, retrieval-augmented generation, and large language models. The work evaluated LLM behaviors on extracted text and documented design choices for pipeline components. Deliverables included research notes and a technical explanation of LLM mechanisms.

Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship

What I've accomplished

I explained LLM token-by-token generation, described how transformers provide contextual processing across tokens, and summarized iterative neural network training used to improve prediction accuracy.

Project breakdown

I described that LLMs generate text by predicting tokens sequentially, explained how transformers provide context by processing multiple tokens at once, and noted that neural networks are trained iteratively to improve prediction accuracy.

View all work