.png&w=256&q=75)
1
1
1 year of experience
Innovative and passionate up-and-coming Data Scientist with 2 years of experience at redBus building high-scale backend systems. Expert in Erlang/OTP and DSA, I specialize in optimizing high-volume data flows and system efficiency. Currently pursuing an MSc in Data Science & AI at Middlesex University Dubai, I am leveraging a strong engineering foundation to master predictive modeling and advanced analytics.
.png&w=828&q=75)
This Partner Churn Predictor & Insights Engine is a dual-model AI framework designed to identify high-risk business partners, quantify churn probability, and deliver automated retention strategies. By synthesizing engagement metrics, financial performance, and support sentiment, the system transforms raw data into a proactive lifecycle management tool. Core Functionalities The project bridges the gap between predictive foresight and granular data retrieval through two specialized components: 1. Predictive Risk Model (The "Forecaster") Utilizing a Random Forest Classifier, this engine evaluates the likelihood of a partner churning. Instead of a binary output, it generates a Risk Probability Score, allowing teams to prioritize accounts by threat severity. Feature Engineering: Processes 12+ variables, including login frequency, referral trends, payout delays, and commission disputes. Contextual Output: Maps each prediction to a specific Churn Driver (e.g., Support Friction) and a Recommended Action (e.g., Fast-track tickets) to ensure interventions are relevant and timely. 2. Partner Insights Model (The "Retriever") A high-speed retrieval system optimized for Account Managers. By inputting a unique Partner_ID, the system instantly pulls a comprehensive profile of historical behavior and current health metrics, enabling deep-dives into the "why" behind risk scores without manual data mining. Technical Stack Machine Learning: Scikit-Learn (Random Forest) Data Processing: Pandas, NumPy Model Serialization: Joblib (for .pkl deployment) Encoding: Label Encoding for multi-region and tiered partner types. Business Impact Traditional churn analysis is often reactive. This engine focuses on the "At-Risk" window, providing the exact probability and specific remedies needed to stabilize a partnership before the loss occurs.
7 Feb 2026