Swadha Tandon

Mettler Toledo - Data Analyst

• Root cause analysis to identify patterns and prevent high- or critical-level IT tickets in a dataset of 1.5 lakh records spanning four years.
• Preprocessed and analyzed ticket data to calculate inter-arrival times (IAT) and generate window-based clustering for CAT (high/critical) and Non-CAT events.
• Identified key recurring patterns and triggers for critical tickets using association rule mining and metrics like support, confidence, and lift.
• Created visual insights (Excel dashboards, color-coded highlights) and presented actionable findings to stakeholders to reduce high/critical tickets.

Tapestry

• Conducted feature-level analysis of product designs using probability & causal analysis. Delivered insights on design attributes impacting sales, supporting data-driven product improvement
• · Leveraged LLMs and AWS services to extract structured features (e.g., Style ID, dimensions, material, color, hardware) from unstructured handbag design sample cards, focusing exclusively on women’s handbags.
• · Built a feature-rich dataset for each design card and performed feature combination analysis using conditional probability to uncover which design features contributed to high or low sales.
• · Delivered clear, data-backed insights to the client showing which features drive sales and which combinations underperform, enabling informed design improvement strategies.
• · Collaborated with teammates on Naive Bayes-based feature prediction and DoWhy analysis to uncover causality and correlation between features and sales outcomes

CRS KM (Internal Quality Bot)

• Worked on scraping of both structured and tabular data for PDF, PPT, docx, Excel to TXT as well as CSV files.
• Training the bot with multiple engines like ChatGPT, GPT4
• Refining the bot instructions to increase the accuracy of the bot as well as enable it to collate data from various files to give a relevant answer with references to the files used.

EA SPORTS

• Analyzed the EA database to identify revenue drivers, sales trends, and competitor growth using clustering & SQL-based queries. Built interactive visual insights to help business teams identify factors affecting revenue.
• Automated analysis of various datasets from the EA database to extract concise insights related to revenue drivers, sales trends, usage patterns, and competitor growth. This involved building an HTML tree-based visual representation of user queries passed and respective insights and decision trees, which will help us identify the components leading to specific metric cohorts. I worked on clustering based on selected metrics like revenue, based on a filter like country, to investigate probable levers affecting it as well.
• Contributing to enhancing the current SQL agent to support multiple tables. This required a thorough study of the client codebase for the current model, which could only support single-table queries, and building the logic to fetch the relationship between different tables from the lambda executor. This critical component enables complex queries involving joins with reduced model hallucinations by curated prompt instructions, thereby improving the precision.
• Beyond this, I rigorously tested the model by simulating various client scenarios throughout the project.

Wellmark - Prompt Engineer

• Prompt Development and Optimization: Crafted and refined prompts for Code Whisperer, tailored to meet diverse client use cases such as code migration, testing, and code conversion. Ensured prompts were efficient and effective in addressing specific technical requirements.
• HTML Code Generation for Angular UI: Created detailed HTML code for Angular-based user interfaces, following specific design templates. Implemented complex UI components, including radio buttons and column layouts, to deliver user-friendly and responsive web applications.
• Exploration and Innovation: Continuously explored various types of prompts to discover new ways to enhance functionality and streamline processes. Focused on improving prompt accuracy and adaptability to support a wide range of programming scenarios.