• Developed forecasting models for "Mother of Pearl," a terrazzo product, using time series analysis and machine learning techniques, contributing to enhanced inventory management and strategic decision-making.• Extracted and analyzed large, complex datasets to identify demand trends, employing Python and SQL for data querying, cleaning, and feature engineering.• Utilized Recursive Feature Elimination (RFE) and correlation thresholds for feature selection, ensuring robust model performance and generalizability.• Built and optimized multivariate forecasting models (including Multilayer Perceptron, Random Forest, LightGBM, CatBoost, and Gradient Boosting Machines) and univariate models (ARIMA, SARIMA, SARIMAX) through hyperparameter tuning, achieving a Mean Absolute Percentage Error (MAPE) of 9.16% (Development), 16.31% (In-time), and 10.30% (Out-of-time) with the optimized CatBoost model.• Collaborating closely with cross-functional teams to deploy models in a real-time system, improving efficiency and accuracy in predicting demand and managing inventory.

Tutoring undergraduate athletes at Columbia College in "MATHV1101 - CALCULUS I", "COMSW1004 - INTRODUCTION-COMPUTER SCIENCE/PROGRAMMING IN JAVA" and "STATGU4001: INTRO TO PROBABILITY & STATISTICS".

• Developed an end-to-end Natural Language Generation (NLG) solution to interpret graphs and tables, enhancing automation in textual analytics• Designed a pipeline to extract, infer, and summarize tables from PDFs using object detection models (TableTransformer), OCR tools, and LLMs.• Built a graph processing pipeline leveraging Google T5, ChartGemma, MiniLM, and MiniCPM models for topic modeling and query resolution.• Automated data extraction from unstructured PDFs, combining contextual text with visual data to generate actionable insights.• Enhanced user experience by integrating topic-specific, personalized narratives into the NLG framework.• Achieved objectives such as minimizing diagnostic and solution identification time and expanding action coverage for diverse scenarios.

• Proposed a novel dense optical flow-based Computer Vision model to track cloth wrinkles• Employed Python libraries (PyTorch, TensorFlow and Keras) to build model• Created a 200GB video dataset, of 128 clothes wrinkling animations, using Blender and MeshLab• Currently working on drafting an IEEE research paper

• Developed an innovative JavaFX vulnerability management system to analyze cyber threats found across four pillars (ROLB, GOLF, DSS, and AS3) of Barclays's infrastructure• Integrated Veracode APIs to detect and filter vulnerabilities into four levels of impact• Automated a Java-based emailing system to send weekly canned reports to entitled personnel

• Led a team of five Toastmasters in club administration, meeting management, communication and reporting of world's biggest Toastmasters Club, with more than 150 members

CRY NGO(Child Rights and You) and TISS (Tata Institute of Social Sciences) have collaborated to conduct a nation wide survey of children between the ages of 9-17, by asking them various questions regarding their Covid-19 experiences. My role is to survey underprivileged children and also lead a team of 10-15 volunteers, who will also be individually surveying the children.

During this internship, I provided valuable inputs to the various online courses conducted by the company. Courses ranging from Data Science, Finance, HR, Public Speaking etc. In this internship I learnt a lot by going through their courses and understanding various concepts.