Hi there 👋

About Me

I’m Emmanuel Sengendo, a University of Michigan-trained Data Scientist passionate about Machine Learning, Artificial Intelligence, and solving complex business problems through data. Based in the beautiful San Francisco Bay Area, I bring technical expertise and creative problem-solving to every project.

What I’m Working On

🔭 Developing a customer service AI assistant powered by state-of-the-art LLMs with optimized context handling and domain-specific fine-tuning
🌱 Advancing my skills in the LangChain framework for building sophisticated AI applications
🏥 Recently completed a spatial-aware hospital recommendation system using neural collaborative filtering and geospatial analysis
🧠 Exploring emerging techniques in multi-modal machine learning and retrieval-augmented generation

Technical Expertise

Machine Learning: PyTorch, TensorFlow, scikit-learn, deep learning architectures
Data Engineering: SQL, ETL pipelines, data preprocessing, feature engineering
Cloud & MLOps: AWS
Languages & Frameworks: Python (advanced), SQL, LangChain, Pandas, NumPy
Visualization: Plotly, Matplotlib, Tableau, geospatial mapping
Natural Language Processing: Transformers, sentiment analysis, text classification

Let’s Connect

I’m open to collaboration on data science projects, particularly those involving NLP, recommendation systems, or healthcare applications.

📫 Email: esengendo@gmail.com
🔗 LinkedIn: LinkedIn
💻 GitHub: GitHub

Education

Master of Applied Data Science

University of Michigan at Ann Arbor

B.S., Business Administration - Information Technology

California State University - East Bay

Projects

Spatial-Aware Hospital Recommendation System

Developed a spatial-aware deep learning recommendation system that helps patients find hospitals matching their care priorities and location preferences.
Engineered custom PyTorch neural network architecture achieving 99.95% accuracy (R² score) by combining collaborative filtering with geospatial features.
Processed and enhanced HCAHPS dataset covering 4,780 healthcare facilities with geocoding, distance metrics, and hospital density features.
Created interactive visualizations revealing geographical patterns in hospital quality and demonstrating recommendation results based on user preferences.
Implemented production-ready recommendation algorithm that balances healthcare quality with practical location constraints for real-world patient decision support. link

End-To-End Text-To-SQL Querying Application

Engineered “IntelliQuery,” integrating Google Palm and LangChain with a custom MySQL database (Threadz E-Store) and Streamlit, for advanced, intuitive database query, and data extraction.
Utilized Few Shot Prompting to optimize SQL database querying and data retrieval, enhancing user experience and efficiency.
Streamlined access to essential data insights via Streamlit-interggrated Text-to-SQL web page, simplifying complex data operations. link

Multi-PDF Document Query and Response Application

Built a PDF Conversational Interface Application integrating LangChain, OpenAI, and Streamlit, enabling interactive document engagement through a conversational chat interface.
Implemented text extraction and advanced processing techniques to segment and vectorize PDF content for efficient information retrieval.
Engineered a responsive Q&A system utilizing conversational models and memory buffers to deliver precise information from PDFs to users.
Designed a user-friendly Streamlit web interface for easy PDF upload, question posing, and instant response viewing. link

IT Helpdesk Ticket Time Prediction Analysis

Crafted an ML model leveraging NLP to predict IT ticket resolution times, enhancing efficiency and customer satisfaction.
Employed text summarization and sentiment analysis for feature extraction to boost predictive accuracy. link

San Francisco Business Closure Analysis Project

Engineered a predictive XGBoost Classifier to forecast San Francisco business closures using historical datasets, targeting key economic and demographic predictors.
Leveraged data science methodologies, including geospatial analytics and clustering, to map and interpret business viability across San Francisco, providing data-driven recommendations for economic strategy and policy development. link

Emotion Detector Image Classification App

Implemented a comprehensive image preprocessing pipeline and leveraged TensorFlow’s Sequential model architecture for facial expression analysis in diverse human emotions. I then deployed the classification web application using TensorFlow and Streamlit. link

Nvidia Stock Price Analysis

Implemented LSTM networks and MACD signals to analyze and predict NVIDIA Corporation stock movements using historical data.
Developed a machine learning model to visualize and forecast stock trends, focusing on educational and exploratory analysis.
Provided an interactive Streamlit application for users to engage with and explore stock price predictions and technical indicators. link