Project Overview
Bidirectional LSTM trained to 85.3% validation accuracy on the IMDB 50K dataset, deployed as a RESTful FastAPI service (POST /predict, GET /health) with a live public Streamlit demo.
Project_Type
ML_API_SERVICE
Entry_Year
2025
Target_Platform
WEB_API
Primary_Role
ML_ENGINEER
Deployed_Technologies
Color_Palette
Project Walkthrough
A closer look at the finer details of this case study. Explore the core features, thoughtful design decisions, and interactive moments that shape the user experience.
[ Page_Index ]
- 01 // The Challenge
- 02 // Interactive Gallery
- 03 // The Solution
01 // The Challenge
The Challenge
Most sentiment-analysis demos stop at a notebook accuracy score and never become something another service can actually call. The goal was to take a trained model all the way to a real, documented, publicly reachable API — with health checks, predictable latency, and a way for non-technical reviewers to try it without writing a single line of code.
02 // Interactive Gallery
03 // The Solution
The Solution
Trained a bidirectional LSTM to 85.3% validation accuracy on the IMDB 50K dataset, then wrapped it in a FastAPI service with POST /predict and GET /health endpoints, deployed behind versioned, documented routes. A companion Streamlit app gives anyone a live, no-code way to type text in and see the model's confidence in real time.
Key_Contributions
Trained and tuned a bidirectional LSTM to 85.3% validation accuracy on the IMDB 50K dataset, then wrapped it as a RESTful FastAPI service with health-check and prediction endpoints, plus a public Streamlit demo for live testing.
Next_Project
PDF RAG Chatbot