My name is
Radhika
I am a Data Science Graduate Student with 2 years of professional experience for Bank of America (TCS).
Available to work Full-time starting May 2025!
My name is
❮
✨ Featured >>>
MACHINE LEARNING & GEN AI Projects:
Predicting Internet Addiction Risk for Mental Health Safety
The Problem: Kids glued to screens? Time to predict future mental health risks!
The Solution: Built AI models (LightGBM, CatBoost, XGBoost) to predict internet overuse with 85% accuracy.
How It Works: Used time-series smart health devices data to cut training time by 30%.
Key Insights: Found the link between physical activity and screen time.
Impact: Made the model easier to understand, predicting future risks long before!
AI Financial Advisor:
Real time Portfolio Recommendation
The Problem: Investors struggle with timely, data-driven portfolio rebalancing.
The Solution: Built an AI tool using AWS and XGBoost to analyze real-time market data.
How It Works: Gives personalized buy/sell recommendations through a fun, interactive dashboard.
Impact: Improved portfolio accuracy by 75% and automated market updates.
Result: Delivered smart investment insights to 1,000+ synthetic client portfolios!
The Problem: Confusing medical jargon? Struggling to find the right doctor?
Healthcare shouldn’t be a puzzle!
The Solution: Built a multilingual AI chatbot using LLMs, RAG, and ranking algorithms
to match patients with the right specialists.
How It Works: Translates complex reports, understands multiple languages,
and delivers clear, personalized recommendations.
Impact: Made specialist access 81% easier, analyzed 10,000+ patient records,
and turned medical confusion into clarity!
Welcome to AI - Risk Detective!
A BERT (NLP) powered risk classifier, which extracts risk factors from reports that are related to AI and classifies them according to their various categories.Tech Stack: NLP, BERT, Pandas, Python
.png)
Check out: Pixel to Plate
A deep learning and computer vision based recipe recommendation system!Key Tech: Machine Learning, Pytorch, Computer Vision, Recommendation System, NLP, Deep Learning, Keras, Paddle
Research Work
Hallucination Detection in Large Language Models
In a groundbreaking endeavor, I got this amazing opportunity to collaborate with the talent researchers at SDSU under Dr. Luwen Huangfu , which aims to develop a state-of-the-art discriminator capable of accurately detecting hallucinations in responses generated by large language models (LLMs).
Leveraging cutting-edge techniques, including transformers, PyTorch, and accelerate, along with powerful 72B training models like LLaMa2 and MoMo, this robust system minimizes cross-entropy loss and ensures the trustworthiness of LLM outputs. Through a comprehensive analysis of hallucination types, employing LLM-assessment metrics, machine evaluations, human assessments, and composite metrics, I aim to unveil invaluable insights into this critical challenge. This pioneering work represents a significant stride towards the responsible and ethical deployment of LLMs, safeguarding against the propagation of misinformation and falsehoods.
Leveraging cutting-edge techniques, including transformers, PyTorch, and accelerate, along with powerful 72B training models like LLaMa2 and MoMo, this robust system minimizes cross-entropy loss and ensures the trustworthiness of LLM outputs. Through a comprehensive analysis of hallucination types, employing LLM-assessment metrics, machine evaluations, human assessments, and composite metrics, I aim to unveil invaluable insights into this critical challenge. This pioneering work represents a significant stride towards the responsible and ethical deployment of LLMs, safeguarding against the propagation of misinformation and falsehoods.
SOME MORE DATA PROJECTS:
Customer Segmentation for Newly Launched Products: Azure ML
Unlocking precision marketing, targeted customers & max profits!
Developed and deployed a powerful K-Means Clustering model on Azure, optimizing feature extraction and hyperparamter selection to launch specific group of products to targeted customers, thereby generating max revenue and more customers.
Data Rx: Cracking the Code on Health Insights
Let's turn numbers into healthier futures!
Into the fascinating world of healthcare analytics where data-driven insights are changing the game, imagine predicting diseases like diabetes, cardiovascular issues, and cancer before they strike - that's what we're up to! With cool scatter plots, heatmaps, and even some map magic using arcGIS, we're decoding the secrets hidden in big health data.
Sentiment Analysis
on Twitter during Covid-19
Unlocking Insights: Advanced NLP-Based COVID-19 Sentiment Analysis
In this technical exploration,by leveraging Natural Language Processing (NLP) and sophisticated data analytics, I tried to decode the public sentiments amidst the pandemic chaos. Our real-time insights from social media data drive agile responses, ensuring decisions and support are precisely aligned with evolving needs for effective crisis management.
Airbnb Price Trends
using Tableau
In this project, I have crafted interactive visualizations that unveil the secrets behind rental rates, from trendy neighborhoods to quirky amenities. Explore the rhythm of booking seasons, uncovering when prices soar and when deals abound. With colorful dashboards, hosts and guests alike can navigate the rental market with ease, turning every stay into an adventure filled with savings and surprises!
Library Management System
Powered by Python, DBMS, and Tkinter. Picture this: admins effortlessly handling inventory, tracking borrowing history, and whipping up reports, all while enjoying a slick interface. And for customers? Think easy-peasy browsing, searching, and reserving books. It's a game-changer for libraries, making things smoother and way more fun for everyone involved!
What's Next?!
MORE PROJECTS COMING SOOON!