Frontend:
React + TypeScript

We chose React for its speed and component-driven flexibility, and TypeScript for its static typing and better dev-time safety. The dashboard interface is fully responsive and designed to give brand users real-time access to key metrics—sentiment trends, keyword insights, and competitor comparisons.

Key Features:

• Filtered search by brand name and timeframe
• Data visualizations (charts, word clouds, trend lines)
• CSV/PDF export for reports

Backend:
Flask + PostgreSQL

Our backend is built with Flask, a lightweight but powerful Python web framework. Flask lets us spin up secure REST APIs quickly and plug directly into Python’s robust data ecosystem.


Data is stored in PostgreSQL, with early prototypes also using SQLite for simplicity. As we scale, PostgreSQL gives us more power for complex queries, joins, and time-series performance.

Auth & Session Management:

• JWT-based token auth (24-hour session tokens)
• Passwords stored with bcrypt hashing
• Role-based access for future multi-brand accounts

NLP: Hugging Face Transformers + PyTorch

For sentiment analysis and keyword modeling, we use pre-trained models from Hugging Face Transformers, particularly BERT-based models fine-tuned for social media text.

Why Hugging Face?

• State-of-the-art NLP
• Flexible pipeline integration via PyTorch
• Support for text classification, token-level tasks

Auth & Session Management:

• Sentiment scores
• High-frequency keywords and phrases
• Emerging topic clusters for brands