Vedika Srivastava

Tech Stack/Skills: Machine Learning, Image Recognition, Neural Networks (NN), Scale-Invariant Feature Transform (SIFT), Docker Containerization, VGG16 Model, Python Programming, GPT-4 Prompt Engineering, Data Analysis, Geographic Information Systems (GIS), API Integration (Mapbox, Google Maps), Cloud Computing, Image Processing, Software Development Best Practices, Team Collaboration

Details:

• Implemented a machine learning-based solution to identify Earth locations in astronaut photos for the Windows on Earth program, significantly enhancing image recognition and geolocation accuracy.
• Utilized Neural Networks (NN) with VGG16 model and Scale-Invariant Feature Transform (SIFT) for robust image analysis and geolocation, overcoming challenges like cloud cover and varied image scales.
• Employed a GPT-4 based AI model, TerraByte, to analyze ISS images, successfully identifying prominent geographical features and providing detailed regional geography.
• Achieved a high success rate in accurately matching ISS images to geographical features, contributing significantly to the fields of environmental monitoring and global mapping.

Tech Stack/Skills: Python, NLP, Rasa Framework, API Integration, Machine Learning, Data Analysis, Security and Privacy, Domain Knowledge (Finance and Stock Market), Version Control, Deployment & Cloud Platforms, Data Augmentation, Sentiment Analysis Tools

Details:

• Developed a conversational AI chatbot in the investment domain, enabling personalized investment advice, real-time stock data, and portfolio management recommendations.
• Utilized Python and Rasa to build the chatbot, ensuring flexibility and customization to deliver a seamless user experience.
• Integrated Alpaca and Alpha Vantage APIs to provide users with real-time and historical market data, ensuring accurate and up-to-date information for making informed investment decisions.
• Implemented sentiment analysis to gauge user satisfaction and feedback, continuously improving the chatbot's performance and responsiveness.
• Successfully addressed domain-specific challenges such as security concerns, high stock market volatility, and complex financial jargon to create a reliable and user-friendly stock market chatbot.

Tech Stack/Skills: Computer Vision, Deep Learning and Neural Networks, Image Processing, Research and Analysis

Details:

• Conducted a comprehensive comparison and analysis of four style transfer models—Neural Style Transfer (NST), Fast Style Transfer (FST), Color-based Artistic Style Transfer (CBST), and Deep Feature Rotation (DFR) for Multimodal Image Style Transfer.
• Assessed the models based on various metrics, including human evaluation, technical evaluation metrics (FID, LPIPS, SSIM), and time taken for style transfer, to determine their effectiveness in generating high-quality stylized images.
• Found that Neural Style Transfer (NST) excels in artistic quality and style fidelity, while Fast Style Transfer (FST) demonstrates faster inference times, making them strong contenders for style transfer applications.
• Potential for Improvement: Identified Color-based Artistic Style Transfer (CBST) and Deep Feature Rotation (DFR) as models with potential for development and improvement in terms of technical quality, opening avenues for further research.
• Provided valuable insights to guide researchers and practitioners in selecting appropriate style transfer models based on specific requirements, such as speed, artistic quality, or technical quality, and emphasized the need for future work in advancing the field of image transformation applications.

Tech Stack/Skills: Computer Vision Algorithms, CNN, GAN, Volumetric Rendering Techniques, Implicit Shape Representations, Gradient Descent Optimization, Deep Image Synthesis, Image-to-Image Translation, Multi-View Geometry, Model Training and Hyperparameter Tuning

Details:

• Pioneered 3D Text2LIVE, an innovative system that empowers localized editing of 3D models using simple text prompts.
• Unleashed the power of cutting-edge Neural Radiance Fields (NeRF) and Contrastive Language Image Pretraining (CLIP) models to achieve unparalleled creative control over 3D model appearance and style.
• Unveiled a groundbreaking process to seamlessly transform generated edits into stunningly realistic 3D point cloud renderings, unlocking a new realm of possibilities for advanced 3D content creation and design.
• Demonstrated the potential of this revolutionary technology across diverse domains, from revolutionizing CAD-modeling to pushing the boundaries of immersive video games and virtual environments.

Tech Stack/Skills: Data Collection & Cleaning, Statistical analysis, Data Visualization, Database management

Details:

• Conducted extensive data collection and analysis using the DAMION Database from multiple counties, exploring questions related to harsher dispositions, sentencing types, and "trifecta" charges for Black and Latino defendants compared to White defendants.
• Employed advanced data visualization techniques to present findings clearly, showcasing significant racial disparities in sentencing outcomes for drug-related charges, Assault and Battery, and other offenses.
• Utilized statistical methods, including the Chi-Square Test of Independence, to establish a strong correlation between race and sentencing outcomes, highlighting the existence of biased practices within the judicial system.
• Contributed to the mission of reducing injustice and inequality in prosecution proceedings by uncovering and shedding light on potential biases affecting marginalized communities.

Tech Stack/Skills: Computer Vision, TensorFlow, OpenCV, CUDA, YOLO, SSD

Details:

• Developed an ML algorithm for drone detection using Computer Vision techniques like YOLO and SSD.
• Successfully increased the accuracy of the YOLO algorithm by expanding the dataset and achieved almost 100% accuracy in drone detection in images.
• Explored and compared the performance of both YOLO and SSD algorithms for drone detection in videos, achieving significant improvements in precision.
• Worked collaboratively in a team of 2 to tackle various aspects of the project, including building the drone, power budgeting, and drone recognition.
• Contributed to the development of an automated drone defense system and surveillance system to address potential misuse and attacks involving drones.

Tech Stack/Skills: Speech Recognition, NLP, Text-to-Speech Conversion, Web API Integration, Data Retrieval and Processing, User Interface Design for Voice Interactions

Details:

• Developed "Zoya," a virtual voice assistant for smart devices capable of natural language interactions.
• Implemented speech recognition using the "speech_recognition" library to convert voice commands to text.
• Employed the "pyttsx3" library for text-to-speech conversion to interact with users audibly.
• Utilized various Python libraries for performing tasks such as web search, Wikipedia search, weather forecasting, and playing music.
• Integrated features like setting reminders, sending emails, and performing navigation tasks to enhance user productivity.

Tech Stack/Skills: Data Analytics, Linear Regression, Logistic Regression, Decision Trees, K-means Clustering, Data Visualization

Details:

• Developed a churn prediction model using Python, data analytics, and machine learning techniques to help a bank retain its customers more effectively. Explored and analyzed a bank dataset containing 11 parameters, including quantitative and categorical variables.
• Employed various statistical techniques, including logistic regression, decision trees, and random forests, to predict customer churn. Conducted chi-square tests of independence to analyze relationships between categorical variables.
• Utilized data visualization techniques to gain insights into the dataset, identify patterns, and detect outliers. Plotted graphs and charts for better understanding and interpretation of the data.
• Identified the most influential factors contributing to customer churn, such as age, gender, balance, and location. Revealed that credit score, estimated salary, and having a credit card did not significantly impact churn.
• Based on the analysis, proposed actionable recommendations for the company, including age-oriented schemes, incentives for maintaining minimum required balances, and targeted strategies for different customer segments (e.g., students and employed individuals).

Tech Stack/Skills: Microcontroller Programming (C8051F340), Hardware Design and Interfacing, Embedded Systems Development, IR Sensor Implementation, Circuit Design and PCB Layout

Details:

• Developed a Density Based Traffic Signal System using IR sensors and microcontroller.
• Implemented a dynamic and automatic traffic light control system to optimize traffic flow.
• Interfaced IR sensors with a microcontroller to detect traffic density on each road.
• Successfully designed a hardware system with LEDs representing traffic signals for a four-side junction.
• Demonstrated effective traffic management by controlling traffic lights based on real-time traffic conditions.

Tech Stack/Skills: Arduino Programming, Sensor Integration (Ultrasonic Sensor), Servo Motor Control, Electronics Circuit Design, PCB Layout and Simulation

Details:

• Developed a "Smart Rodent Trap" using Arduino and ultrasonic sensors.
• Utilized a servo motor to close the cage and trap the rodent alive.
• Incorporated LEDs and a buzzer to provide visual and audio feedback of successful trapping.
• Designed a cost-effective and efficient solution for humane pest control in rural areas.
• Implemented hardware and software design to create a fully functional rodent trap system.

Tech Stack/Skills: Electronic Circuit Design, IC 555 Timer Configuration, Light Sensor Calibration (LDR), Understanding Photoelectric Devices (Photo Sensors), Troubleshooting and Testing Electronic Circuits

Details:

• Developed a Light Sensor Circuit using LDR and Timer 555 for detecting light intensity.
• Utilized 555 Timer IC to control the buzzer based on LDR resistance variations.
• Implemented a simple and cost-effective solution for automatic light sensing and control.
• Practical applications include solar highway lighting, street light control, and security alarm systems.
• Addressed limitations such as slow response time and temperature dependency of LDRs.

Tech Stack/Skills: Java Programming, Object-Oriented Programming (OOP), GUI (Graphical User Interface) Design, Data Randomization and Handling, Application Development with BlueJ

Details:

• Created a miniature version of the Mindspark online learning platform as a standalone application.
• Implemented a user-friendly interface with a main page and topic selection feature.
• Developed a random question generator to present different questions during each practice session.
• Successfully simulated the learning experience by offering 10 random questions for each chosen topic.
• Utilized Java and BlueJ for application development.