Yaswanth Pidugu

I’m a B.Tech CSE (AI & ML) student with a strong focus on applying machine learning to real, high-impact problems—especially semantic search, recommendations, and forecasting.

In my LeetCode Semantic Recommender System, I architected a hybrid Learning-to-Rank framework using Sentence-BERT and LightGBM, analyzing 3,700+ problems and reaching an NDCG score of 0.60. I also automated data ingestion with a custom GraphQL scraper and engineered a 4-signal ranking feature set that reduced user problem-discovery time by 40% on my deployed Streamlit app.

I’ve also built practical computer-vision pipelines with PESTOPIA, creating a lightweight CNN (1.5M parameters) optimized for edge devices, achieving 92% validation accuracy. To improve robustness, I implemented stratified data augmentation, expanding 1,500 base samples to 13,500+ images and reducing overfitting.

Across projects like Rossmann Retail Sales Forecasting, I focus on measurable results—forecasting 6-week sales trends (R score 0.95) and improving MAE by 15% through hyperparameter tuning. I back my engineering with certifications in Generative AI (Oracle University) and Python essentials (Cisco Networking Academy), and I’m eager to contribute to a team that values strong experimentation, clean pipelines, and real-world deployment.