Skip to main content

AI Systems Engineering Portfolio

End-to-end machine learning systems workflow covering data processing, model training, optimization, deployment, and monitoring with production-grade best practices.

Get StartedView Architecture
✓Reproducible training pipeline
✓ONNX export & quantization
✓Hardware-aware optimization
✓Production FastAPI endpoints
✓Real-time drift monitoring

Quick start

Get up and running with the ML systems workflow in minutes

1

Install dependencies

Clone the repository and install required packages:
git clone https://github.com/RaviTejaMedarametla/Data-Science-AI-Portfolio.git
cd Data-Science-AI-Portfolio
pip install -r requirements.txt
2

Run the training pipeline

Execute the configuration-driven training workflow:
python -m src.train
This will:
  • Load and preprocess data from ml_datasource.csv
  • Engineer features using the configuration in config.yaml
  • Train multiple models with cross-validation
  • Select the best model and save it to artifacts/
  • Generate metrics, lineage, and drift baseline files
3

Start the prediction API

Launch the FastAPI server for real-time predictions:
uvicorn src.api:app --host 0.0.0.0 --port 8000
Access the interactive API documentation at http://localhost:8000/docs
4

Make your first prediction

Send a prediction request to the API:
curl -X POST "http://localhost:8000/predict" \
  -H "Content-Type: application/json" \
  -d '{
    "student_country": "United States",
    "days_on_platform": 45,
    "minutes_watched": 320.5,
    "courses_started": 3,
    "practice_exams_started": 2,
    "practice_exams_passed": 1,
    "minutes_spent_on_exams": 85.0
  }'

{
  "predicted_purchase_probability": 0.82,
  "predicted_purchase": 1
}

Explore the system

Navigate through the complete ML lifecycle

Training

Learn how to train models with reproducible pipelines, feature engineering, and cross-validation

Deployment

Export models to ONNX, apply quantization, and validate numerical parity

Optimization

Benchmark performance, analyze hardware trade-offs, and tune for production

Runtime

Deploy FastAPI endpoints for predictions, QA, and streaming inference

API Reference

Explore the complete API surface with request/response schemas

Case Studies

Real-world applications in healthcare, edge AI, and streaming analytics

Key features

Production-grade capabilities for enterprise ML systems

Configuration-Driven Training

Reproducible model training with YAML configuration, deterministic seeds, and lineage tracking

ONNX Export & Quantization

Convert scikit-learn models to ONNX with INT8 quantization for optimized CPU inference

Statistical Benchmarking

Measure latency, throughput, and accuracy with bootstrap confidence intervals and significance testing

Drift Detection & Monitoring

Track feature drift and prediction rate shifts with automatic retraining triggers

Production API Endpoints

FastAPI services for predictions, batch inference, QA, and streaming with health checks

RAG-Based QA System

Question answering API powered by LangChain, ChromaDB, and OpenAI with streaming support

Ready to explore the system?

Dive into the complete ML lifecycle workflow with hands-on guides and real-world examples

View Quickstart GuideExplore API Reference