Aadvait Hirde

Bachelor of Science: Major in Computer Science, Specialization in AI, Minor in Business

2023 - 2027

Cumulative GPA: 3.81 / 4.0

・Dean's List・Hutton Honors Scholar・Global Engagement Scholar・Luddy Direct Admit

Relevant Coursework: Discrete Math, Calculus I, Data Structures and Algorithms, Artificial Intelligence, Object Oriented Programming, Data Analysis and Mining, Applied Algorithms, Managerial Economics and Strategy, Operations Management

• Founded the IU chapter of KTP, the nation's professional tech fraternity, leading a 15-member executive board and driving the chapter's strategic direction.
• Managed internal operations, including budgeting, marketing, university recognition, and national communication, securing affiliation with both the PFC and KSG.
• Formed strategic partnerships with CS and business clubs and co-developed an interdisciplinary case competition with real companies and campus orgs.
• Built an integrated application and CRM system using Airtable, Google Forms, and Zapier to streamline multi-stage recruitment.

• Developed a scalable recommendation system using Python, TensorFlow, and PyTorch, improving personalized recommendations for over 100,000 users.
• Designed and implemented feature engineering pipelines to process large-scale datasets (5TB+) using Apache Spark and Pandas, reducing preprocessing time by 30%.
• Conducted model performance analysis using precision, recall, F1 score, and ROC-AUC metrics, achieving a ~12% improvement over baseline models.
• Fine-tuned state-of-the-art transformer models (BERT, GPT) for text classification tasks, achieving a 95% accuracy rate in sentiment analysis for user reviews.
• Collaborated with a cross-functional team to deploy machine learning models in production via Flask and FastAPI, utilizing AWS SageMaker for model hosting.

MNIST and CIFAR-based Real-Time Image Classification Systems using CNNs and Tensorflow

Aug 2024 - Nov 2024

CIFAR Code・CIFAR Paper・MNIST Code・MNIST Paper

• Developed real-time image classification systems for both MNIST digit recognition and CIFAR-10 object classification, achieving over 90% accuracy across various object classes.
• Implemented CNNs with regularization techniques to optimize performance and prevent overfitting.
• Created interactive OpenCV interfaces for real-time object and digit recognition from live webcam footage.
• Processed and enhanced images using techniques like Gaussian blurring, binary thresholding, and data augmentation.

Forecasting Cryptocurrency Prices with Transformers and Stance Detection

Jun 2022 - Nov 2022

IEEE Xplore・PDF

• Developed a stance detection model using the RoBERTa transformer, achieving 80% accuracy and MAE of $1144.85 in predicting tweet sentiment towards Bitcoin.
• Utilized Twitter API for retrieving Bitcoin-related tweets and yfinance library for historical price data.
• Applied logistic regression for sentiment analysis, achieving 78% accuracy in predicting tweet sentiment.