ML Engineer · Data Scientist · East Africa

Freddy
Nyanda

I build predictive ML systems and data products that help teams increase retention, grow revenue, and deliver actionable insights at scale across East Africa.

View Projects Get in touch
5+
Models deployed
200M+
Data processed
12+
Projects shipped
4+
Years experience
Freddy Nyanda
Python Expert
AI Solutions
About Me

Building ML that moves East Africa forward

I am an ML Engineer & Data Scientist with 4+ years of experience developing predictive and prescriptive machine learning solutions across mobile money, telecom, and research environments.

My expertise lies in building robust models for customer segmentation, churn prediction, and revenue forecasting. I have hands-on experience deploying LightGBM and XGBoost pipelines into production and translating complex business challenges into actionable AI-driven insights.

Currently, I work as a Pricing & CVM Analyst at Mixx by Yas (Axian Group), where I leverage data to optimize pricing strategies and enhance customer value across Tanzania.

Experience

4+ Years

Specialization

Machine Learning

Location

Dar es Salaam 🇹🇿

What I enjoy building

  • End-to-end ML pipelines — data, model, deployment
  • Customer analytics dashboards & segmentation tools
  • Forecasting & optimization models for revenue & churn
  • Production-grade APIs and automated data products
  • CI/CD pipelines for ML workflows
  • Real-time financial tracking & prediction systems
Work Experience

Where I've delivered impact

Roles focused on turning data into measurable business outcomes using AI-driven systems.

Jan 2025 – Present
Pricing & CVM Analyst
Mixx by Yas · Axian Group
  • Develop predictive churn and revenue forecasting models using Python, LightGBM, and XGBoost to support customer value management.
  • Built 7-segment behavioral customer segmentation model (Super Power User → Passive) to optimize campaign targeting, improve ARPU, and enhance retention.
  • Maintains ML pipelines, ETL procedures, data cubes, and automated reporting systems on SQL Server.
  • Built PushPromo daily prediction pipeline and Luku SMS Reminder Engine running in production.
  • Collaborated with commercial teams to translate business problems into data-driven AI solutions.
Jan 2023 – Dec 2024
Lead Data Scientist
AI4D Research Lab
  • Led end-to-end ML lifecycle, from data acquisition and EDA to model development, deployment, and monitoring.
  • Deployed deep learning-based prediction models using TensorFlow/PyTorch for early-risk identification.
  • Improved model accuracy and stability through systematic feature engineering and tuning.
  • Worked with multidisciplinary teams to deliver AI-driven insights for operational and policy planning.
Skills & Competencies

Tools I use to build & deploy

A snapshot of the tools and techniques I use to design, build, and deploy data products.

🧠 ML & Analytics

LightGBM XGBoost Scikit-learn Predictive Modeling Customer Segmentation Time Series Feature Engineering EDA

🗄️ Data Platforms

SQL Server Python Pandas NumPy ETL Pipelines Power BI Tableau REST APIs

⚙️ Engineering

Docker GitHub Actions CI/CD FastAPI Streamlit Git Azure AI Linux
Featured Projects

Real-world ML systems

Production ML and data science projects with automated pipelines and business impact.

Strait of Hormuz Monitor

Strait of Hormuz Monitor

Real-time AIS vessel tracking for the Strait of Hormuz closure, cross-referenced against live Brent crude prices. Persistent ingestion runs on a self-hosted cloud VM (websocket tracking needs an always-on process, not a scheduled job) with a live public dashboard.

AIS/WebSockets SQLite Self-hosted Cron Oracle Cloud
Code Live Demo
EA Financial Tracker

EA Financial Tracker

Real-time East Africa exchange rates & crypto tracker with ML predictions. Self-hosted cron pipeline (crypto every 30 min, FX hourly) for reliable sub-hourly updates. Streamlit dashboard with live charts and forecasts.

Streamlit Scikit-learn Self-hosted Cron
Code Live Demo
Global Fuel Watch

Global Fuel Watch

Fuel price intelligence across 20 countries — petrol, diesel, LPG, and crude oil. Split-cadence pipeline (FX every 20 min, crude every 4 hours) designed around a strict 25-request/day API quota discovered during build-out.

Pandas Self-hosted Cron API Design
Code
East Africa News Sentiment

East Africa News Sentiment

NLP sentiment tracking for Kenyan, Tanzanian, Ugandan, Rwandan, and Burundian news. Rebuilt on a per-country-verified source mix after discovering mainstream news APIs' free tiers don't actually cover these countries — verified real regional sources before switching providers.

VADER NLP Streamlit Self-hosted Cron
Code Live Demo
Tanzania Climate Watch

Tanzania Climate Watch

Weather monitoring and ML forecasting across 5 Tanzanian cities with z-score anomaly detection, running every 15 minutes on self-hosted cron. Caught genuine pressure anomalies in Dar es Salaam and Zanzibar during testing.

Scikit-learn Self-hosted Cron Anomaly Detection
Code
Credit Card Segmentation

Credit Card Segmentation

Customer segmentation using clustering and dimensionality reduction to identify high-value and at-risk cohorts. Includes interactive dashboards for segment profiling and targeting.

Python Scikit-learn K-Means Pandas
Code Notebook
DSN Anomaly Tracker

DSN Anomaly Tracker

Automated monitoring pipeline for NASA's Deep Space Network, ingesting live feed data, engineering features, and scoring anomalies on a self-hosted cron schedule with a live operations dashboard.

Pandas Matplotlib Self-hosted Cron
Code Live Demo
Customer Churn Prediction

Customer Churn Prediction

End-to-end churn prediction pipeline using Logistic Regression, Random Forest, and XGBoost on transaction data. Includes model monitoring dashboards and feature importance analysis.

Python XGBoost Scikit-learn Dash
Code Notebook
Car Price Predictor Tanzania

Car Price Predictor — Tanzania 🇹🇿

ML-powered used car price predictor with real-time TZS conversion. LightGBM model trained on 19,000+ cars achieving R²=0.74. Full CI/CD pipeline with automated testing and FastAPI deployment.

LightGBM FastAPI CI/CD Python
Code
Contact

Get in touch

Interested in collaborating or have a project in mind? Send a message and I'll get back to you.

Need a quicker response? Reach me directly at freddynyanda@proton.me