🧠
Mind Reader
An end-to-end text intelligence platform that compares classical ML and Transformer models in real time to surface sentiment, classification, and feature insights.
🟣
Overview
Mind Reader is a scalable ML platform that analyzes text through multiple models—classical ML, Transformers, and custom pipelines—to generate insights such as sentiment, classification, and feature importance.
🌱
Problem
Text data is rich, but hard to process at scale. Researchers, students, and teams often need a simple tool to compare models and visualize results instantly.
🧩
What I Built
- •End-to-end ML pipeline with 4 different models
- •Real-time comparison between classical ML (SVM) and large Transformer models (BERT)
- •TF-IDF + Linear SVM feature engineering workflow
- •Hyperparameter tuning with GridSearchCV + cross-validation
- •Streamlit dashboard for interactive exploration
- •Visualization for decision boundaries, class distributions, and model outputs
⚙️
Technical Deep Dive
Models
SVM, BERT, TensorFlow NLP models
Pipeline
Hugging Face, NumPy, TF-IDF, GridSearchCV
Frontend
Streamlit (interactive UI + real-time processing)
Deployment
Docker container → Kubernetes cluster
Accuracy
Achieved 92% on sentiment analysis tasks
🌈
Impact
- •Enables fast iteration across different NLP approaches
- •Makes model comparison intuitive for non-technical users
- •Fully productionized pipeline with containerized deployment
👩🏻💻
My Role
Machine learning engineer, data scientist, frontend builder, and deployment owner.