Trupti Shriyan

Developed a Random Forest Regression model to predict KSAT from 40+ unstructured Excel soil datasets, achieving an R2 of 80% and reduced RMSLE by applying extensive feature engineering and cross-sample evaluation, demonstrating model scalability across diverse geotechnical data.

Fine-tuned a 12-layer RoBERTa transformer on 74K+ text samples using 3-fold cross-validation, achieving 91.2% accuracy and 91.1% F1-score by merging 4 public datasets into a balanced MetaHate corpus, applying byte-level BPE tokenization, and accelerating training by 40% using FP16 precision on A100 GPUs.

Built a multi-model classification pipeline using Random Forest, SVM, Logistic Regression, kNN, and Nave Bayes, achieving 95% accuracy on the Wisconsin Diagnostic dataset by applying Chi-square feature selection, scaling, hyperparameter tuning, and visualizing results with heatmaps and confusion matrices for enhanced interpretability.

Developed a multiclass classification model using One-vs-Rest logistic regression, achieving 85% accuracy and a 20% performance gain over baseline by applying cross-validation, hyperparameter tuning, and result visualization with Matplotlib and Seaborn. Collaborated with business stakeholders to define success metrics, align modeling goals with operational objectives, and translate technical resul

Currently pursuing a Master of Science in Computer Science, focusing on advanced topics in the field. Expected to graduate in May 2026.

Grade: 3.75 / 4.00

Completed a Bachelor of Technology in Computer Engineering, gaining a strong foundation in computer science principles. Achieved a GPA of 3.75 out of 4.00.