|
MSE in Data Science @ University of Pennsylvania
Data Science (AWS) RA @ Wharton • Research in AI & ML • Turning Data into Impactful Decisions
About Me
Hi, I'm Chaitanya Kakade, currently pursuing my Master of Science in Engineering (M.S.E.) in Data Science at the University of Pennsylvania (UPenn). I graduated with a Bachelor of Engineering (B.E.) in Computer Engineering from the University of Mumbai.
My passion lies in the intersection of AI technologies, specifically in machine learning and computer vision. I'm especially drawn to creating solutions that don't just push boundaries technically but also carry tangible impact.
My research journey has taken me through some exciting milestones, from contributing to IEEE conferences in domains like Computer Vision, NLP, Geoscience, and Remote Sensing, to leading projects that aim to solve real-world problems using AI. These experiences have deepened my understanding and strengthened my commitment to uncovering how AI technologies can drive innovation, inspire solutions, and shape our evolving digital landscape.
Along the way, I've become skilled in working with complex datasets, designing data pipelines, and building efficient models. Currently, my focus is on data-centric AI research and the design of production-grade machine learning systems. I'm also actively exploring emerging developments in Agentic AI, multimodal learning, and their integration into business applications.
Key Achievements
Let's Connect
Technical Expertise
Programming Languages
ML/AI Frameworks
Cloud Platforms
Tools & Platforms
Work Experience
Data Science (AWS) Graduate Research Assistant
The Wharton School
Graduate Research Assistant focused on AWS-based data science projects
Key Achievements
- Coming soon...
Research Collaboration
Tata Consultancy Services (TCS)
Collaborated with Dr. Shailesh Deshpande, Principal Scientist at TCS Research
Key Achievements
- Implemented a novel image segmentation workflow in QGIS using Python scripts on Landsat Satellite data, cutting analysis time by 60%
- Produced more accurate burn severity maps estimating 25.8M tonnes of carbon emissions
- Showcased research at IEEE IGARSS 2024 in Athens, captivating over 500 academics and industry experts
- Generated 12 follow-up conversations regarding potential workflow adoption
Data Science Intern
Uniconverge Technologies Pvt. Ltd.
Focused on IoT sensor data analysis and fault detection systems
Key Achievements
- Conducted EDA on 2M+ IOT sensor readings, extracted actionable insights
- Identified anomaly patterns and fault signatures in gearbox machinery system
- Achieved 15% improvement in early fault detection
- Engineered 20+ statistical features and validated SVM based fault classification models
- Achieved 92% accuracy leading to reduced downtime across 10+ monitored machines
Data Science Intern
PHN Technologies Pvt. Ltd.
AWS cloud-based data processing and predictive modeling
Key Achievements
- Processed 10,000+ records from AWS S3 using Python, Pandas, SQL
- Performed feature engineering to improve data usability, reducing missing/incorrect data by 25%
- Developed and tested an XGBoost model in AWS SageMaker, achieving 92% predictive accuracy
- Designed interactive dashboards in Tableau and AWS QuickSight
- Reduced report preparation time by 30%
Featured Projects
Virtual Try-On: AI-Driven Fashion Technology
Developed a 2D upper-body virtual try-on pipeline using diffusion-based inpainting with Stable Diffusion, SAM (Segment Anything Model) for precise garment masking, and advanced image processing techniques. Addressed the critical challenge of online fashion returns (30-40%) by enabling customers to virtually try on clothing. Implemented pose keypoint alignment, thin-plate-spline warping, and ControlNet for structural guidance to achieve realistic garment fitting while preserving identity, pose, and background.
Border Surveillance System with AI-Driven Thermal Vision
Developed an AI-driven border surveillance system using thermal and night vision with modified Faster R-CNN architecture. Ranked Top 30 of 5,000+ teams at DIPEX 2025 and secured 1st position at National Project Exhibition.
DocBot - Disease Prediction System
Built a comprehensive disease prediction system using machine learning on medical symptoms dataset. Achieved 97.3% accuracy with multiple classification models deployed on AWS Cloud.
Small Language Model Implementation with Resource Optimization
Built a small language model using PyTorch, trained and tested on the Encyclopedia Britannica and TinyStories datasets. Used a BPE tokenizer, efficient binary storage, a 6-layer Transformer with multi-head attention, and mixed precision training on M3 Pro GPU. Achieved stable training with AdamW and cosine annealing, generating coherent text outputs.
Advanced Image Reconstruction Autoencoders
Developed a convolutional autoencoder for image reconstruction, achieving 0.92 SSIM and 38.5 dB PSNR. Implemented multi-stage convolutional and upsampling layers, reducing dimensionality by 75% while preserving 90% visual information. Conducted comparative analysis across epochs, optimizing reconstruction error from 0.0156 to 0.0021.
A Minimal MCP Client & Server Demo
A simple repository that shows the process of building an MCP server and using Claude Desktop as a client. Features a Travel Desk system to handle employee travel requests, approvals, and history tracking — all accessible directly from Claude. Demonstrates how to modify the contents to develop specific MCP use cases.
Llama-2-7B-GGML-Powered Blog Generator
Using the advanced Llama 2 7B Chat model by Meta, this project offers a seamless experience for generating high-quality blogs with just a few clicks. Features AI-powered blog generation, customizable writing styles (Fun, General, Professional), word count specification, and a user-friendly Streamlit-based web interface.
Online Retail Analysis using Fireducks
Comprehensive analysis of online retail data demonstrating how Fireducks significantly speeds up data processing and analysis compared to traditional methods. Showcases performance improvements in data manipulation, aggregation, and visualization for retail analytics.
High Impact Research
Advancing the frontiers of AI, Machine Learning, and Data Science through innovative research
A Multimodal Framework for Spatiotemporal Causal Analysis of Mumbai's Air Pollution Using Social Media Insights and Remote Sensing
This paper presents a novel multimodal framework for analyzing air pollution patterns in Mumbai using spatiotemporal causal analysis techniques, combining satellite data with ground-based measurements and social media insights.
Optimal Detection of Diabetic Retinopathy Severity Levels Using Attention-Based CNN and Vision Transformers (ViT)
Diagnosing diabetic retinopathy (DR) from colour fundus images is challenging as it requires skilled clinicians to recognize and interpret multiple small features. This study presents an approach for detecting the severity levels of diabetic retinopathy using a combination of Attention-Based Convolutional Neural Networks (CNN) and Vision Transformers (ViT).
Effectiveness of Kolmogorov-Arnold Networks (KANs) and Analysis of Machine Learning Algorithms in Heart Disease Prediction
This research explores the effectiveness of Kolmogorov-Arnold Networks (KANs) in heart disease prediction compared to traditional machine learning algorithms, providing insights into the potential of this novel neural network architecture in medical diagnosis.
Evaluating Real-NVP For Spatio-Temporal Modelling In Synthetic Soil Nutrient Data Generation
This study evaluates the effectiveness of Real-NVP (Real-valued Non-Volume Preserving) transformations for spatio-temporal modeling in synthetic soil nutrient data generation, exploring applications in precision agriculture and environmental monitoring.
Enhancing Sign Language Interpretation with Multi-Headed CNN, Hand Landmarks and Large Language Model (LLM)
Sign language is an important mode of communication for the deaf and mute community. Despite its importance, there is still a large communication gap between deaf community and the hearing world. We introduce a new system that converts sign language into text, using a novel multiheaded Convolutional Neural Network (CNN) that is trained on three different sets of images-raw images, uniquely segmented images, and hand landmarks information simultaneously to recognize sign language gestures accurately for different hand textures under disparate background conditions. Additionally, a Large Language Model (LLM) is incorporated to transform these recognized signs into concise and meaningful sentences.
Carbon Emission Estimation in Sahyadri (Western Ghats) Resulting from Burning Grassland Biomass
Biomass burning contributes to large quantities of gaseous pollutants and aerosol particles in the atmosphere, having a significant impact on air quality, human health, and climate. This study examines the complex topic of estimating carbon emissions from grassland biomass burning in the Sahyadri region (Western Ghats). We calculate total burnt area over the period of ~5 months using normalized burned index and innovative methods of subsequent subtraction technique.
Predictive Analytics for Enhancing Crop Yield using Generative Adversarial Networks and its Challenges
In the contemporary landscape of agriculture, predictive analytics has emerged as a pivotal tool for enhancing crop yields. The integration of Generative Adversarial Networks (GANs) into this domain presents a novel approach that promises to significantly improve predictive accuracy. This research delves into the application of GANs for crop yield prediction, highlighting both the potential benefits and inherent challenges.
MotionScript: Sign Language to Voice Converter
Sign language serves as a vital mode of communication for the deaf and mute community, yet it presents a significant barrier in their interaction with the larger society, which often lacks proficiency in sign language. This paper presents MotionScript, an innovative sign language to voice conversion system that leverages computer vision, deep learning, Convolutional Neural Networks (CNN), Natural Language Processing (NLP) and Large Language Model (LLM) to facilitate interaction between individuals from the deaf and mute community and the rest of the world. This paper outlines a thorough comparison of four distinct neural network models, utilizing metrics to identify the most accurate model for transforming American Sign Language (ASL) into coherent and meaningful sentences voiced in natural language.
Research Impact
Contributing to the advancement of AI and Data Science
Loading articles...
Awards & Recognition
Celebrating achievements in academic excellence, research competitions, and professional speaking engagements
Principal's Excellence Award
Awarded to the top 5% of students for academic and research excellence
DIPEX 2025 Project Exhibition - Top 30
Ranked Top 30 of 5,000+ teams at DIPEX 2025 with a functional prototype, presented to investors and India's Defence Research and Development Organisation (DRDO)
U' LECTRO '25 National Level Project Expo - 1st Position
Secured 1st Position in the AI/ML Domain
Guest Speaker – Deep Learning Workshop
Led a 3-hour session on Deep Learning, covering neural networks, backpropagation, optimizers, and advanced topics including LLMs and Transformers for real-world AI applications
Guest Speaker – QGIS & Machine Learning for Research
Delivered a session on applying QGIS and ML concepts to solve real-world problems and support research initiatives
Invited Speaker - Geospatial Computing and Applications
Delivered an in-depth demonstration of real-time data visualization and analytics from IoT sensors suspended in a water body using QGIS, demonstrating temperature patterns and insights
Invited Speaker - Workshop on Analyzing Vegetation Health with ML
Demonstrated practical applications of Machine Learning, Geospatial data, and Remote Sensing, showcasing live examples of how to use satellite imagery and tools like Google Earth Engine and QGIS to address real-world challenges
ResCon 2024, Research Presentation Competition - IIT Bombay
Secured 3rd place among 100+ teams
Techno Kagaz 2024, Research Conference
2nd Runner Up
Get In Touch
Let's discuss your data science needs or collaborate on exciting projects
Let's Connect
I'm always interested in hearing about new opportunities, interesting projects, or just having a chat about data science.
kakadechaitanya77@gmail.com
Phone
(267) 258-6268
Location
Philadelphia, PA