Raunak Gangwal

I’m a Computer Engineering graduate focused on turning messy data into reliable insights and building ML/GenAI features that users can actually use. I combine strong data analysis foundations with practical model development and deployment using modern tooling.

In my Customer Churn Analysis & Prediction work, I built an end-to-end churn prediction pipeline on 7,043 telecom customers. I used Python, data cleaning, EDA, feature engineering, and tuned Logistic Regression and Random Forest models, achieving 0.846 ROC-AUC and 79% recall, while generating a ranked retention list for 518 high-risk customers and approximately $503K/year in at-risk revenue.

I also built an AI Code Explainer platform using LangChain and Meta’s Llama 3.3 70B via the NVIDIA API. The system supports multiple languages with plain-English explanations, line-by-line comments, and complexity analysis, and I implemented bug detection and improvement recommendations using a FastAPI backend and a React + Tailwind CSS frontend.

On the research side, I co-authored a published Dual-Domain DeepFake Detection framework and developed a FastAPI inference pipeline. Using a ConvNeXtV2-Tiny backbone with FFT-based frequency feature extraction, staged freeze–unfreeze training, and learnable gated feature fusion, I reported 97.70% AUC on FaceForensics++ and 96.39% AUC on Celeb-DF.