Arnav Kucheriya

• Engineered machine learning models for real-time solar radio burst detection, enhancing space weather prediction.
• Preprocessed large datasets with Pandas, NumPy, and Scikit-learn, ensuring data readiness and model reliability.
• Validated predictive models and contributed to research publications, utilizing TensorFlow and PyTorch for optimized analytics.

• Collaborated with the Central Bank of Oman (CBO) to deploy an ML-driven HR recruitment system, reducing screening time by 50%.
• Designed an automation pipeline for Vietnam's State Bank, improving transaction efficiency by 30%.
• Built a REST API using FastAPI and PostgreSQL to streamline data storage and ML model integrations.

• Automated ETL processes with Python, Django, and PostgreSQL, reducing processing time by 40%.
• Synchronized multi-platform databases, ensuring 100% data accuracy and integrity.
• Integrated user data with marketing platforms (Keap, Hi-Sawyer, Wix), increasing campaign targeting efficiency.

GPT-Scratch: Building Character-Level Language Models

• Designed and implemented two character-level language models—Bigram and Mini GPT—from scratch using PyTorch.
• Trained models on Shakespearean text to analyze loss, perplexity, and GPU usage, comparing statistical vs. transformer-based approaches.
• Developed custom embedding layers, self-attention heads, and transformer blocks to enable contextual text generation.
• Showcased significant improvements in coherence and fluency with the transformer model over the Bigram baseline.

View on GitHub | Project Paper

Market Prediction with Machine Learning

• Developed a financial modeling and analysis tool to predict stock prices using machine learning algorithms.
• Utilized Python libraries like Pandas and NumPy for data preprocessing and feature engineering.
• Implemented and optimized models such as linear regression and decision trees to analyze market trends.
• Visualized results using Matplotlib to deliver actionable insights for investment strategies.

View on GitHub

Google Workspace Chatbot

• Developed and trained an AI-powered chatbot hosted on Streamlit to handle real-time query processing and automate IT help desk tasks.
• Designed the chatbot to continuously learn from training data, providing efficient and accurate responses.
• Improved response time and reduced workload for IT staff by replacing manual support operations.

View on GitHub

Unity Game Development

• Design and develop advanced games and systems using the Unity Game engine.
• Created and implemented a Ray-Tracing System by building a Data Pipeline to Label and Correct Data.
• Built interactive games optimized for both Mobile Application and WebGL platforms.

View on GitHub