Kelvin Mwaura

Dedicated data scientist with hands-on experience in machine learning, algorithm development, and end-to-end model deployment. Skilled at transforming complex datasets into actionable insights through rigorous exploratory analysis, feature engineering, and predictive modelling using frameworks such as TensorFlow and scikit-learn. Proven ability to ship ML-powered applications on Streamlit, Flask, and Django, and to scale workloads on AWS. Also fluent in BI storytelling through Power BI and Tableau dashboards, enabling businesses to gain clear and actionable insights from their data.

These are some of my personal projects that have found real case use.

Car Price Prediction

Moringa School · Aug 2024 · github.com/TyrionCodister/car_price-prediction

Built an end-to-end ML pipeline to solve pricing inefficiencies in Kenya’s used car market, performing feature engineering on vehicle attributes (age, mileage, brand, condition) and training a model with R² of 0.80 and low MAE, enabling accurate, data-driven vehicle valuations

FlickFlare Movie Recommendation System

Moringa School · Jul 2024 · github.com/TyrionCodister/FlickFlare--Movie--Recommendation--System

FlickFlare, an innovative movie streaming platform, offers a wide array of movies, from timeless classics to the latest blockbusters. To achieve user satisfaction, FlickFlare wanted to develop a personalized recommendation system addressing user concerns about content relevance.

For this project, I implemented a sophisticated recommendation engine leveraging collaborative filtering techniques (SVD, NMF, and KNN) to deliver personalized top 5 movie suggestions based on users' past ratings and genre preferences. The recommendation system aims to increase user engagement, watch time, and retention, setting FlickFlare apart in the competitive streaming market.

SyriaTel Customer Churn PredictionMi

Syria Tel · Jun 2024 · github.com/TyrionCodister/Churn-in-Telecom-

Syriatel, a leading telecommunications company, faced a significant customer churn challenge, impacting profitability and increasing the costs of new customer acquisition. In the highly competitive telecom industry, retaining existing customers is crucial for maintaining market position and revenue.

In this project, I developed a predictive model to identify customers most likely to churn, enabling Syriatel to implement proactive retention strategies. By analyzing key indicators of customer behavior and building a classification model, I provided actionable insights and recommendations to help Syriatel reduce churn rates and strengthen customer loyalty.