My Customer Churn Prediction & Risk Tier Engine uses Python, SQL, and Scikit-Learn to identify churn risk in bank accounts. I built a leakage-safe pipeline that achieved a 0.77 PR-AUC in five-fold cross-validation, compared with 0.69 for a logistic regression baseline.
For TrailQuest AI, I built an offline-first app that generates structured outdoor quests. I configured Ollama with Gemma 2B and Qwen 2.5B for local inference, and submitted the project to the Touch Grass Open-Source AI Challenge.
At SSRC, I completed data science training and applied it in hands-on data cleaning, analysis, and dashboard projects. I also created mathematics content at Mathematics Initiatives in Nepal (MIN) and taught robotics at SARC Education Foundation.
