Designed and implemented a full Databricks Medallion Architecture (Bronze → Silver → Gold) pipeline to extract, cleanse, and normalize unstructured invoice data using Spark, Delta Lake, and advanced regex-based parsing.

Engineered automated product categorization logic and built analytical models to compute revenue, VAT, quantity trends, and year-over-year metrics, improving data usability and business insight generation.

Developed an interactive Databricks SQL dashboard featuring category distribution, timeline trends, and gross/net comparisons, enabling clear visualization of key financial and operational insights.

Implemented an emotion recognition system integrating deep learning models for synchronized text and visual inputs.

Preprocessed speech and visual frames using OpenCV and Whisper, applying semantic and exact text-to-emotion mapping.

Fine-tuned Florence-2 for emotion classification, achieving 69% accuracy and a training loss of 1.67, while optimizing GPU utilization, batch processing, and memory management for efficient model training.

Conducted a global study on life expectancy disparities using WHO and UN data, analyzing key socioeconomic, healthcare, and environmental factors across countries.

Performed data cleaning, handled missing values, and conducted EDA with correlation heatmaps, box plots, and geographic visualizations.

Developed ML models (Logistic, Ridge, Lasso Regression); the best model achieved 87% accuracy and identified income composition and child mortality as top predictors.

Applied predictive modeling and healthcare analytics to a real-world public health challenge, strengthening expertise in R, data visualization, and regression techniques.

Built a simulated live cricket analytics dashboard using historical match data and backend API-driven updates to mimic real-time performance.

Processed structured datasets for key metrics (scoring rate, partnerships, player stats) and applied data cleaning, transformation, and dynamic aggregation.

Visualized match dynamics through line charts (run rate trends), bar plots (top batsmen), partnership graphs, and team comparison donut charts using R, ggplot2, and Quarto.

Revealed strong correlation between top partnerships and total scores, with visual insights into team dominance and scoring fluctuations enhancing real-time sports analytics skills.

Designed a Power BI dashboard to visualize visitor trends to India, integrating multiple Kaggle datasets covering foreign tourists, overseas Indians, and crew demographics.

Performed data cleaning, reshaping, and merging in R using JOIN operations; created conceptual, logical, and physical data models for efficient schema design.

Implemented DAX measures for dynamic aggregations (e.g., total visitors, year-over-year change, gender ratios) and built demographic filters for age group, gender, country, and mode of travel.

Developed interactive visuals including line graphs, tornado plots, treemaps, and stacked charts to analyze visitor trends and demographics, uncovering key patterns across age, country, and entry points.

Designed a playable Checkers game with human vs. human and human vs. AI modes, implementing core mechanics like move validation, piece capturing, and winner detection.

Built a Monte Carlo simulation–based AI in Python that evaluates random future game outcomes to select the most strategic move.

Overcame challenges related to game state modeling, randomness control, and AI integration through iterative debugging and testing.

Demonstrated strong AI performance in simulations, with future plans to integrate reinforcement learning for smarter opponent move anticipation.

Analyzed grade outcomes (DEW rates) across semesters pre-, mid-, and post-COVID to uncover patterns in student performance using DEW and course attribute datasets.

Applied EDA, anomaly detection, and time series analysis, revealing that fully online, asynchronous courses had notably higher DEW rates—especially during Fall 2020.

Built decision tree and regression models to identify key predictors such as course modality and session timing, though overall predictive accuracy remained limited.

Highlighted the evolving impact of online learning and proposed future enhancements through expanded data sources and refined modeling techniques.

Developed a custom GPT-like NLP system to generate responses from university datasets using cosine similarity, Excel-based input, and tools like Python, Scikit-learn, NLTK, and OpenPyXL.

Preprocessed institutional documents and fine-tuned the model for context-aware answers; evaluated accuracy through expert validation.

Built a user-friendly interface with HTML/CSS/JS and Node.js backend; modeled system behavior with UML diagrams (class, sequence, activity).

Conducted extensive testing (unit, integration, regression) and proposed future enhancements including sentiment analysis and cross-institution NLP applications.

Applied full data science lifecycle using IBM Watson Studio and Cloud Pak for Data — including data cleaning, EDA, modeling, and visualization — to solve a real-world problem.

Performed feature engineering and built supervised (regression, classification) and unsupervised (K-Means, DBSCAN) models to uncover insights and make predictions.

Evaluated models using accuracy, precision-recall, AUC-ROC, and RMSE; used AutoAI to automate model selection and tuning for improved performance.

Communicated business insights through dashboards, Python-based visualizations, and data storytelling to guide data-driven decision-making.