StudentPass AI driven Student Performance Management

StudentPass: Predicting Student Performance (Pass/Fail)

StudentPass is an ML-based web application that predicts whether a student will pass or fail based on academic and demographic data. It helps educators and institutions quickly identify at-risk students and take timely interventions. StudentPass reduces manual effort, speeds up performance evaluation, and provides actionable insights using predictive machine learning models.

Here’s a walkthrough of what I built:

Phase 1 – Data Collection

End goal (business perspective):

1. Automate student performance prediction using machine learning.
2. Reduce manual effort for educators while providing actionable insights. 3.Collect student data with a binary target variable (pass/fail) and understand feature importance.
3. Understand requirements for an API and batch prediction workflow.

Phase 2 – Data Ingestion with MongoDB

Goal: Store and manage collected data in a NoSQL database.

Tools Used:

1. MongoDB for data persistence

2. PyMongo for Python–MongoDB interaction

Key Steps:

Create database and collection in MongoDB

Insert raw or cleaned data for later access

Phase 3 – Model Development

Goal: Train and evaluate a machine learning model.

Tools Used:

1. scikit-learn for ML

2. pandas and numpy for preprocessing

Outcome: A trained logistic regression model stored as logreg.pkl.

Phase 4 – Flask API

Goal: Serve the ML model through a REST API.

Tools Used:

1. Flask for API endpoints

2. Flask-PyMongo to fetch input data from MongoDB

Features:

Single predictions via JSON input

Batch predictions via CSV input/output

Phase 5 – Dockerization

Goal: Containerize the application for consistent deployment.

Tools Used:

1. Docker for containerization

2. docker-compose to run multiple services

Key Files:

Dockerfile

docker-compose.yml

Phase 6 – Monitoring with Prometheus

Goal: Collect application and system metrics.

Tools Used:

Prometheus for metrics scraping and storage

node-exporter for host CPU/memory metrics

cAdvisor for container-level metrics

Phase 7 – Visualization with Grafana

Goal: Visualize metrics and build dashboards.

Tools Used:

Grafana for visualization

Imported Prometheus as a data source

Dashboards: CPU, memory, and API request metrics

Key Takeaways

Built an end-to-end ML pipeline from scratch

Containerized services for consistent deployment

Implemented monitoring and visualization for production readiness

Learned how modern ML engineering combines coding, DevOps, and system observability

Tech Stack

Tool/Library	Purpose
Python	Data processing, model training
pandas, numpy	Data wrangling
scikit-learn	ML model training
MongoDB	Data storage
PyMongo	Python–MongoDB connection
Flask	Serve model as API
Docker	Containerization
Prometheus	Metrics collection
node-exporter	Host system metrics
cAdvisor	Container-level metrics
Grafana	Visualization dashboards