Writing
Technical notes & fiction
Deep-dives on the ML pipelines I ship, and a serialized story running on the side — two different registers, kept clearly apart.
01Technical writing
End To End Machine Learning Project-Part I
10 min readDesigned and implemented a full-lifecycle machine learning pipeline to classify toxic online comments. The solution integrates datasets from Kaggle and YouTube, addressing severe class imbalance through Synthetic Minority Over-sampling Technique (SMOTE). Leveraged TF-IDF for feature extraction and trained a custom Neural Network using TensorFlow/Keras, successfully serializing the model for deployment readiness.
Data Science · Jun 25, 2024
LLM Text Detection
23 min readArchitected a full-stack MLOps solution to distinguish between human and AI-generated text. The project features a modular production pipeline (Ingestion to Evaluation), a custom TensorFlow Neural Network with TF-IDF vectorization, and is fully deployed on AWS EC2 with a Streamlit interface for real-time inference.
Data Science · Jul 11, 2024
End-To-End ML Project: Spam Mail Classification
10 min readMoving from simple prompts to autonomous agents. Exploring ReAct patterns, tool use, and memory management for building capable AI assistants.
Data Science · Jul 15, 2024
