Shresth Bhargava
Back to Projects
Backend Engineering

Vellum

AutoResearch AI Startup Platform

Rebuilt the backend of an AI-powered startup documentation platform from Python/FastAPI into Java Spring Boot with production-oriented architecture.

Java 21Spring BootPostgreSQLJWT
Architecture
Layered

Controller-Service-Repository

Auth
Stateless JWT

BCrypt & UUID identifiers

AI Integration
RAG Pipeline

Multi-agent generation

Migration from Python to Java

The original prototype was built in Python (FastAPI). As the platform scaled to handle complex multi-agent document generation, maintaining type safety and robust enterprise patterns became critical. I led the complete rewrite into a Java Spring Boot monolith.

Layered Architecture

I enforced a strict Controller-Service-Repository pattern with constructor-based dependency injection. This decoupled our business logic from HTTP concerns, allowing us to easily write unit tests using JUnit 5 and Mockito, while keeping the API layer thin and maintainable.

RAG Pipeline for Business Documents

Currently developing a RAG-based multi-agent pipeline to generate BRDs (Business Requirement Documents) and market research. The Spring Boot backend securely manages the vector database connections, ensuring AI models are grounded with user-specific contextual data to prevent hallucinations.