Researchers have developed a scalable online machine learning system for stock recommendations, addressing the challenges of market volatility and low-latency predictions. The system utilizes a distributed microservices architecture with Kubernetes, Docker, and RabbitMQ, featuring a leader-follower model for continuous training on streaming financial data and parallel serving of recommendations. Implemented with TensorFlow Recommenders, it generates content-based recommendations and ensures high availability through persistent model weights and message durability. The architecture demonstrates efficient performance, serving recommendations in 23 seconds per request and handling up to 500 portfolio additions per second per follower, though it faces limitations with data staleness due to API rate limits. AI
IMPACT Provides a production-ready reference architecture for online recommender systems in finance, balancing scalability and availability.
RANK_REASON The cluster contains a research paper detailing a new system architecture for stock recommendations. [lever_c_demoted from research: ic=1 ai=0.7]
- Alpha Vantage API
- Docker
- Google Cloud Object Storage
- Harsh Ketan Nagarkar
- Kubernetes
- RabbitMQ
- TensorFlow Recommenders
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
