Prathyusha Murala
AI & Data Science Strategist
Bridging the gap between cutting-edge AI methodologies and real-world business outcomes, I specialize in building intelligent, data-driven solutions that unlock strategic value. My expertise encompasses the entire data science lifecycle—from advanced predictive modeling, forecasting, and personalized recommendations to natural language processing and deep learning—ensuring each solution is explainable, ethically grounded, and closely aligned with organizational objectives.
Professional Experience
My journey in Data Science and AI
Associate - Data Scientist
PricewaterhouseCoopers
Led the design and deployment of an AI-powered compliance automation system for vendor security risk assessments, tailored to ISO/IEC 27001 standards. Applied transformer-based NLP models (BERT, SBERT, KeyBERT) to extract and classify security clauses from unstructured documents with high precision.
Enhanced classification accuracy by 40% and reduced manual review cycles by 60%, enabling analysts to complete risk evaluations in under 4 hours (previously 3 days). This initiative streamlined onboarding across global business units, resulting in $500K in annual savings and faster regulatory alignment.
Financial Data Analyst - Intern
Wipro
Built and deployed data-driven tools to support financial decision-making, using Python (Django, Plotly) and NSEpy to process over 20 years of historical market data. Developed real-time analytics pipelines that enabled rapid trend detection and anomaly tracking in stock behaviors.
Delivered a 45% boost in real-time processing speed and reduced data defect rates by 35%, enhancing the accuracy of algorithmic trading insights. Empowered analysts to conduct faster market simulations, resulting in a 15% improvement in forecasting efficiency during portfolio analysis.
Educational Background
Syracuse University
Master of Science - Applied Data Science
Relevant Coursework: Quantitative Reasoning for Data Science, Responsible AI, Information Visualization, Natural Language Processing
Dayananda Sagar University
Bachelor of Technology - Computer Science and Engineering
Relevant Coursework: Machine Learning, Deep Learning, Data Warehousing & Data Mining, DBMS
Projects
Explore my work!
EEG-Based Emotion Recognition with LSTM + DWT
Developed a deep learning pipeline to classify emotions using EEG signals. Combined Power Spectral Density (PSD) and Discrete Wavelet Transform (DWT) features with an LSTM architecture, achieving 96.76% accuracy, outperforming SVM, KNN, and RF baselines.
- Architecture: LSTM + DWT Feature Extraction
- Comparative Models: SVM (95.98%), RF (91.15%), KNN (93.65%)
- Tools: TensorFlow, NumPy, SciPy
- Published: ERCICA Conference, Springer 2023
LexiSense: Semantic Book Recommender with SBERT
Built a hybrid recommender using SBERT embeddings, clustering, and collaborative filtering to enhance personalization and discovery.
- Embeddings: DistilBERT, T5, SBERT
- Techniques: UMAP + K-Means (0.6 Silhouette Score)
- Collaborative Layer: SVD-based filtering
- Improvement: 20% boost in recommendation accuracy
World Happiness Project
Explored multi-year World Happiness Report & OECD data to identify top factors influencing happiness across 50+ countries.
- Dataset: 5+ years of country-level happiness, income, and mental health data
- Model/Technique: Trend analysis, interactive Tableau dashboard
- Tools/Platform: Tableau, Python, OECD data API
- Result: Linked social investments to a 15% rise in happiness; guided policy decisions
UCI Adult Dataset Analysis
Predicted income brackets using demographic and occupational features, with an emphasis on model fairness and interpretability.
- Dataset: UCI Adult Dataset (45K+ records)
- Model/Technique: Logistic Regression, Random Forest; fairness metrics, LIME, SHAP
- Tools/Platform: Python, scikit-learn, SHAP, LIME
- Result: Identified model bias and improved interpretability for responsible AI deployment
Harmony Hub Music Streaming Service
Designed a scalable database and analytics solution for music streaming data, enabling real-time insights into user listening habits and genre trends.
- Dataset: Platform data of tracks, users, albums, and streaming sessions across US states
- Model/Technique: Normalized relational schema, SQL queries, triggers, procedures
- Tools/Platform: SQL, ER diagrams
- Result: Enabled advanced analytics on user engagement and subscription optimization
Hierarchical Text Classification with Attention Mechanisms
Developed a novel hierarchical text classification framework that leverages attention mechanisms to improve document categorization accuracy across complex taxonomies.
- Innovation: Multi-level attention for hierarchical classification
- Dataset: Reuters RCV1, arXiv papers
- Results: 8.3% improvement in F1-score over existing methods
- Published: ACL Workshop on NLP Applications, 2023
Skills & Expertise
My technical toolbox built from industry experience and academic excellence
Contact
Let's connect! Whether you're interested in discussing potential opportunities, have a collaboration in mind, or just want to talk data science, feel free to reach out. You can connect with me on LinkedIn or Email. I look forward to hearing from you!