Sonita Nayak

Dynamic Demand Forecasting:

• Forecasting models were developed to predict demand for laptops, printers, and accessories across regions and sales channels.
• Multiple data sources—including historical sales, promotional calendars, and external events—were integrated to enhance prediction accuracy.
• Machine learning techniques such as XGBoost, Prophet, and LSTM were employed to capture seasonal and promotional effects.
• Models were adapted for both consumer and enterprise product lines, accounting for variations in buying behavior and lead times.

Recommendation System for Personalized Banking Products

• Developed a hybrid recommendation engine to personalize retail banking product suggestions (credit cards, loans, insurance, savings plans) for over 500K+ active users.
• Combined collaborative filtering (LightFM) with customer demographics, transactional behavior, and product metadata to address cold-start and data sparsity challenges.
• Built a feature-rich training pipeline with customer segmentation, RFM metrics, and NLP-based product embeddings (TF-IDF + PCA) to enhance content-based predictions.
• Improved product uptake by 17% in targeted email/SMS campaigns (A/B tested), driving measurable revenue increase in cross-sell and upsell efforts.
• Collaborated with marketing, compliance, and product teams to ensure model explainability and adherence to fairness guidelines in financial recommendations.
• Deployed the model as a REST API using FastAPI, integrated with the bank’s CRM and mobile app for real-time customer interactions.
• Evaluated model performance using Precision@K, MAP, and uplift metrics; created dashboards in Power BI for stakeholder visibility.

• Reporting at accounts level data, Automated monthly tableau reports.
• Analyzing customer verbatims to learn about the ongoing customer issues and give appropriate recommendations to business.
• Built data pipelines using python, PySpark to aggregate data across varied data sources.

• Customer Segmentation - Clustering customers-based on their usage of banking services/applications.
• EDA on unstructured data, text processing / text mining.
• Scatter plots to understand the data distribution.
• Used K-means algorithm and evaluation using inertia and silhouette score.
• NLP using NLTK, LDA, Gensim covering regular expression, feature extraction and topic modeling.

• Oracle ERP Techno Functional Consultant in Finance Domain Making payments, processing AP AR invoices. End to End Implementation.
• Advanced SQL queries.
• Development of high-quality database solutions using Python and SQL.