Bebeto Nyamwamu

- Developed a FastAPI-based platform for geospatial location evaluations, handling thousands of monthly requests on AWS EC2. - Integrated AWS Cognito for secure user authentication and access control, enabling scalable and role-based access management. - Implemented AWS ElastiCache Redis for caching geospatial H3 hexagon data, reducing query latency by 40% for dynamic viewport analytics. - Designed CI/CD Bitbucket Pipelines for Dockerized services, cutting deployment time.

- Built batched ETL pipelines on GCP Compute Engine, ingesting and transforming 500K+ survey records monthly into BigQuery for analytics. - Orchestrated data workflows with Prefect and Docker, ensuring reliability and 99% uptime across production pipelines. - Developed and optimized data models and schemas in BigQuery to support performant queries and scalable analytics. - Built Looker dashboards on top of BigQuery, enabling 40+ stakeholders to access real-time market insights with a 95% adoption rate.

Data Processing, Behaviour Analysis and Modelling for Healthcare Delivery Optimization - Tech Stack & Skills: Python - Data Processing Pipelines using pySpark for healthcare delivery optimization - Cleaning, Feature Engineering, and Analysis of partners' data using pySpark, Azure Databricks, seaborn, matplotlib - Time Series, Survival Analysis, and Modelling to understand and predict user churn and time to associated events using pySurvival and Lifelines with over at least 60% model accuracies

Tech: Python, Tensorflow Developing an AI Model to Identify School Locations in Sudan from Satellite Imagery

Tech: Tensorflow | Geopandas | Numpy Using Satellite Imagery to Assess FAW - Collaborated in a team of 50 to collect relevant data - Collaborated in a team to preprocess satellite images - Built a Resnet-50 model to predict the area affected by Armyworm pest and classify level of infestation - Accuracy of over 90% and RMSE of 0.36

Tech: XGBoost, Scikit-Learn, Streamlit - Cleaned data - Processed data to fit machine learning - Classified startups (closure, operating, ipo or acquisition) with an accuracy of over 89%. - Startup Recommender App up and running. Link: https://start-upp.herokuapp.com

Managed the task of collecting, cleaning and organizing data from online sources for Biofoods, hotels and restaurants.

- Extract data from SQL and NoSQL databases - Clean authentic, messy datasets - Build predictive models for regression and classification - Create compelling visualizations to explain data characteristics and model performance. - Discuss the ethical implications of deploying models in the real world. - Consider the environmental impact of machine learning models. - Learn how to apply machine learning to business problems.