A Leader, Data Analyst and Team Player.
My Projects
Welcome to my portfolio. Here you’ll get to know me through the work I do. Feel free to look around!
Enhancing Financial Analysis with Generative AI.
Machine Learning LLM's
The 5-month project with a FAANG company utilizes techniques such as one-shot prompting, query rephrasing, and context window adjustment to improve the accuracy of financial metric extraction from SEC 10-k Documents.
We examined the potential of Generative Artificial Intelligence (Gen AI) for improving financial analysis by automatically extracting metrics through a combination of Large Language Models (LLMs), Natural Language Processing (NLP), and other techniques.
The project demonstrates the accuracy of the Text-Bison 32k from 16% to 58% through a combination of techniques and model selection. This is supplemented with a chatbot deployable in the financial space.
Addressing the Housing Crisis in Tippecanoe County.
Business Acumen Market Research Consulting
Led a team to devise pragmatic solutions to tackle the urgent need for affordable housing in the Greater Lafayette Area comprising of 6 counties. Developed methods to combat affordable housing via focusing on LIHTC, ADU's, Eviction Prevention and affordable modular housing. Presented in front of two NGO's: Homestead CS and Habitat for Humanity. Placed 1'st position in Purdue.
NCAA Women's Basketball Ticket Sales Prediction.
Machine Learning Visualization Spreadsheets
Competed with 67 teams with 280 students participating in the Crossroads Classic Analytics Challenge. Displayed a classification accuracy of 98.68%, ranking us 1st at Purdue and 2nd nationally along with the best visualization award.
We were challenged with understanding customer purchasing behavior for the Division I Women's Basketball Tournament. After analyzing a dataset of 200,000 records with 25 features we generated additional features along with machine learning techniques to improve prediction accuracy. Along with this, we built advanced machine learning algorithms like Logistic Regression, XGBoost, and Linear SVC were used. Experimentation with ensemble techniques further enhanced prediction accuracy.
We employed machine learning models, created Tableau dashboards to present our findings, and delivered presentations. A/B testing was also utilized to experiment with various sales strategies. Proposed the introduction of loyalty programs to incentivize primary market purchases and boost customer retention.
Profiling Neurology KOL's for Product Launch Success.
Machine Learning Data Analysis Statistical Analysis
The project chalenge aimed to identify and profile Key Opinion Leaders (KOLs) in neurology to support the product launch for a client in the US neurology pharmaceutical market. The approach involved a robust data-driven analysis using multiple sources to assess KOLs based on various metrics such as total payments, clinical trials, publications, and their respective impacts within the medical community.
The data sources included public databases and rankings related to neurology and neurosurgery. Advanced analytics techniques, like the Analytic Hierarchy Process (AHP), were employed to prioritize and weigh different metrics, creating comprehensive profiles that included an "Impact Score" for each KOL. This score combined contributions in publications, clinical trials, and leadership to evaluate their overall influence and potential impact for the product launch.
Leveraging Data to Help Members Become "Super"
Machine Learning LLM's Streamlit
This project aims to enhance Airbnb listing performance by analyzing the dynamics of achieving Super host status. Utilizing advanced machine learning models and a rich panel dataset, we provide insights to improve guest experiences and boost overall platform success.
The comprehensive analysis and strategic recommendations aim to significantly improve host performance and guest satisfaction, reinforcing Airbnb's position in the market.
Pricing Strategy Analysis for New Restaurant Entrants.
Pricing Strategy Statistical Modeling Data Analysis
Our project focused on analyzing the Indian food delivery market, specifically Swiggy, to offer insights for new restaurant entrants. We utilized a comprehensive Kaggle dataset to examine customer preferences, price sensitivity, and the impact of delivery performance. We analyzed key aspects like cuisine types, delivery times, number of ratings, and price brackets through various statistical methods including one-hot encoding and part-worth analysis.
Our aim was to empower restaurant owners to make data-driven decisions tailored to customer expectations and market dynamics. We visualized the insights using Tableau, providing an interactive way to explore the factors that influence customer choices across different cities.
Optimizing Healthcare Entity Extraction Using LLM's.
Machine Learning LLM's Healthcare
The project aimed to improve healthcare entity extraction using large language models (LLMs), focusing on automating tasks and ensuring compliance. We analyzed 2,001 transcripts from a Kaggle competition, optimizing with LLMs to lower the Word Error Rate (WER) to 0.53738, achieving 4th place among 95 teams. We selected openrouter.ai for API integration, refining our methods for optimal Named Entity Recognition (NER) outputs.
Our development included coding for transcript queries, error correction, and data cleaning. This work showcased LLMs' potential to enhance healthcare operations efficiently.
A Winning Approach to Firm Collapse Prediction.
Machine Learning Data Mining Finance
In the MGMT 571 in-class competition at Purdue University, our team excelled by finishing 1st on the public leaderboard, 2nd on the private leaderboard, and securing 1st place overall in the competition. The challenge centered on developing a predictive model for firm collapse, using various econometric measures to assess the financial distress and long-term viability of companies. The goal was to predict the probability of bankruptcy for each firm, with the AUC metric serving as the basis for evaluation. Our efforts allowed us to achieve an AUC of 0.95857.
Building and Growing an Online Healthcare Platform.
Digital marketing and Advertising Data Analytics Product Development
Leading Save The Young Heart as the Marketing and Operations Head has been a transformative experience. I oversaw the development and launch of our website, which became a central hub for user engagement and health education. Our team effectively utilized digital marketing strategies across multiple platforms like Google Ads, Facebook, and Quora, significantly expanding our reach to over 800,000 people.
The website not only enhanced our digital presence with over 31,000 unique visitors but also served as a platform for crucial health analytics. Through targeted marketing efforts, we achieved a remarkable increase in revenue by over 104% and improved cost efficiency. This journey was pivotal in promoting coronary artery disease awareness and empowering individuals with tools to assess their heart health risks.
Leveraging LinkedIn Data for Strategic Job Market Insights.
Data Visualization Data Analysis
Our project focused on using LinkedIn data to provide comprehensive insights into the 2023 job market. By analyzing over 15,000 job postings, we aimed to empower job seekers and recruiters with actionable information. Our visual analytics explored job roles, required skills, and salary trends across various industries.
We created interactive Tableau dashboards to display job distribution and salary expectations, enhancing job search efficiency for job seekers. Our visualizations highlighted trends such as job listing popularity by state, job count by company size, and the relationship between work type and industry sponsorship. This in-depth analysis helped job seekers make informed decisions and provided recruiters with valuable market insights.
Elevating Hyderabad House: Strategic Marketing at Purdue
Digital Marketing and Advertising Consumer Behavior Strategy
Our pitch for Hyderabad House involved a targeted marketing campaign to boost its visibility and foot traffic at Purdue University, home to a large Indian student community. We launched a phased advertising approach on Instagram that increased profile visits by over 500%. Our ads, themed around cultural heritage and broader appeal, notably improved engagement, with our culturally-themed ad doubling link clicks.
We utilized sensory marketing to enhance the dining experience, connecting Hyderabad House’s biryani with luxurious imagery and flavors. This approach not only heightened brand perception but also strategically expanded our customer base through ongoing refinement of our marketing tactics.
Gap's Strategic Shift to Data-Driven Fashion Retail
Machine Learning Retail Market Analysis LLM's
Explored Gap Inc.'s potential to enhance its competitive edge by adopting a big data-driven strategy. Our analysis incorporated Google Trends for market trend identification, consumer feedback analysis using GPT-3.5 for sentiment assessment, and competitive analysis via web scraping with Selenium and ChatGPT to monitor fashion trends and product popularity.
Additionally, we conducted regression analyses on sales data from platforms like Amazon and Google Shopping to determine the impact of pricing, discounts, and reviews on product ratings. Our comprehensive approach suggested that integrating big data could significantly refine Gap’s marketing strategies, aligning product offerings with consumer trends and preferences while balancing creative branding with precise, data-driven decision-making.
Optimizing Lucky Store with Database Solutions.
Database
Our project focused on developing a comprehensive database system for the Lucky Store to enhance its operational efficiency and data-driven decision-making. We transitioned the store’s management to a secure, structured database system that integrates Key Performance Indicators (KPIs) for performance monitoring and enables the extraction of actionable insights to boost revenue and support expansion plans.
The database design included tables for items, categories, customers, transactions, payment methods, suppliers, and item sales. This structure supports detailed analyses such as total sales per category, identifying unsold items, and analyzing payment method preferences. Our implementation ensures that the Lucky Store can optimize inventory, refine its store layout, and improve customer engagement through targeted marketing and loyalty programs.
Boiler Bargains Student Couponing at Purdue
Product Development Web Development Strategy
Our presentation detailed the development and impact of "Boiler Bargains," a Purdue-specific couponing website designed to address student needs for affordable shopping options. Key findings from our market survey indicated that a significant majority of students actively seek deals, with 94.5% looking for food coupons and over 70% for clothing discounts. Our platform, BoilerBargains.shop, is tailored to enhance usability and trust among students, featuring Purdue-localized deals and rigorously verified coupons.
With a simple vendor charge model, our site not only simplifies access to discounts but also aligns with student budget constraints. The presentation outlined our phased launch strategy, beginning with a minimum viable product and scaling through user feedback and strategic marketing to maximize reach and utility. The website is currently down.
Pocket Ninjas: AI-Powered Data Dashboards
LLM's Product Development Streamit
Pocket Ninjas was my experimental project aimed at developing AI-integrated data dashboards that allow users to interactively query and understand company data. The project successfully combined data visualization tools with AI capabilities, enabling users to ask questions directly to the data, enhancing the analytical process. The dashboards were designed to be user-friendly and were customizable to fit the specific needs of various clients.
Pocket Ninjas managed to secure several clients, showcasing the potential and effectiveness of integrating AI with data analytics to simplify complex data interactions and provide actionable insights efficiently.
SQL-Driven Fraud Detection for Financial Transactions
Database Management Finance Machine Learning
The project developed a sophisticated SQL-based real-time fraud detection system for a leading financial institution. We addressed the rise in digital financial fraud by utilizing SQL to secure and analyze transaction data, customer profiles, merchant information, and account activities. Our methods included constructing entity-relationship diagrams, developing relational schemas, and extensive data cleaning and normalization.
We implemented advanced SQL queries for identifying anomalies and detecting irregular transaction patterns, significantly enhancing security measures and bolstering customer trust against financial fraud.
