Ammar Saqlain

Built a hybrid product recommendation model combining collaborative filtering and content-based approaches to generate personalized product suggestions. Validated model performance by comparing behavior on training splits versus held-out test data and used Precision@K and Recall@K to assess recommendation quality. Developed an inventory demand prediction model to forecast whether individual products would see increased demand in the following month using XGBoost and LightGBM. Established a baseline using a 4-week lag approach and evaluated ML models against it to support reorder decisions. Presented results through an internal demo dashboard.

Researched and fine-tuned Large Language Models to develop a specialized diagnostic tool for automotive troubleshooting. Evaluated OpenAI APIs to improve response accuracy for complex mechanical queries. Transitioned into frontend integration by working with React and JavaScript, resolving UI/UX bugs and implementing feature fixes to improve interface stability and usability.