Belva AI: Building LLM Microservices Infrastructure
By Xavier Collantes
Created: 3/20/2025; Updated: 7/14/2025
At Belva AI I served as a Software Engineer building robust
microservices infrastructure for AI-powered applications.
Technical Development and API Architecture
During my time at Belva AI, I was instrumental in developing the backend
infrastructure that powered our AI services:
- Developed 25+ REST API endpoints across 10+ different LLM microservices using FastAPI, Nginx, creating a comprehensive and scalable API ecosystem
- Implemented event-driven architecture using Kafka for asynchronous processing and communication between microservices
- Built data persistence layers with MongoDB, optimizing for both performance and flexibility in storing AI-generated content and user data
- Containerized all services using Docker and orchestrated deployments with Kubernetes to ensure consistent environments across development and production
Professional Testimonial
AI/ML Engineer at Belva AI
"we shipped more than 20 AI agents together, and the process was always smooth thanks to his attention to detail and sense of ownership."
- Victor Guang Ming Lee, Ph.D.
Database Architecture and Design
I played a key role in the architectural planning and implementation of database
solutions:
- Designed and architected database backend options for a user-facing AI-driven full-stack tool
- Prepared detailed cost-benefit analyses for each database solution, presenting tradeoffs to the CTO
- Implemented the selected database architecture, ensuring it met performance, scalability, and reliability requirements
- Optimized query patterns and established proper indexing strategies to maintain fast response times even as data volumes grew
Technical Stack and Skills Applied
My work at Belva AI leveraged a modern technology stack including:
- Python with FastAPI for efficient API development
- Ollama for testing LLM models against each other
- Kafka for message streaming and event processing
- MongoDB for flexible document storage
- Docker and Kubernetes for containerization and orchestration
- Microservices architecture design and implementation
- CI/CD pipeline configuration for automated testing and deployment
This experience at Belva AI deepened my expertise in building distributed
systems for AI applications, particularly in designing and implementing scalable
infrastructure that can handle the unique demands of LLM-based services.
Related Articles
Related by topics:
thingsIBuilt
llm
python
api