ML/AI

Recommendation Engine

Collaborative filtering engine for personalized plan suggestions.

Client
Internal Project
Industry
Machine Learning / Healthcare
Duration
3 months (2024)
Role
ML Engineer
Recommendation Engine

Overview

Built a recommendation system that suggests optimal health plans based on member profiles, behavior patterns, and similar user preferences.

Key Results

40%
Higher
Conversion
25%
Less
Plan Switches
<50ms
Latency
Per Request
85%
Precision
@10

The Challenge

Members struggled to find the right health plans:

  • Too many plan options causing decision paralysis
  • Generic recommendations not personalized
  • High plan switch rate due to poor matches
  • No data-driven plan optimization
  • Manual recommendation process

They needed intelligent, personalized recommendations.

The Solution

I built a hybrid recommendation system:

Collaborative Filtering

Matrix factorization to find similar member preferences.

Content-Based Filtering

Plan feature matching based on member profiles.

Behavior Clustering

K-means clustering to group similar members.

Real-Time Scoring

Redis-cached recommendations with fast retrieval.

Key Features

🎯

Personalized

Recommendations tailored to each member.

🔄

Hybrid Model

Combines collaborative and content-based filtering.

âš¡

Real-Time

Instant recommendations with caching.

📊

A/B Testing

Built-in experimentation framework.

Tech Stack

ML

TensorFlowScikit-learnPandas

Backend

FastAPIPython

Data

PostgreSQLRedis

Infrastructure

DockerAWSAirflow

Screenshots

Recs

Recommendations UI

Analytics

Performance Analytics

Model

Model Architecture

Achievements

  • Increased plan conversion by 40%
  • Reduced plan switch rate by 25%
  • Sub-50ms recommendation latency
  • 85% precision@10 in offline evaluation