
Engineer of AI systems specializing in intelligent information retrieval. Currently developing production-ready RAG applications utilizing custom vector databases, with a focus on hybrid search systems that integrate semantic understanding and lexical precision. Recent achievements include a custom vector database using FAISS and a hybrid search engine that merges BGE-Large embeddings with BM25 for an Azure-deployed RAG system in the education sector. Background in data analytics exceeding five years supports expertise in ML evaluation and statistical testing, effectively translating business challenges into technical solutions.
Custom Full-Stack RAG System | GitHub: github.com/RicciJuaman/custom-rag-faiss
Python • React • FAISS • PostgreSQL • FastAPI • Azure • CI/CD
- Built production RAG application with custom FAISS vector database implementing hybrid semantic-lexical search
- Engineered full-stack solution: React frontend with real-time search, FastAPI backend, PostgreSQL metadata layer
- Implemented hybrid retrieval combining BGE-Large embeddings with BM25 scoring for improved accuracy on both
conceptual and exact-match queries
- Developed CI/CD pipeline with automated testing and deployment to Azure infrastructure
- Achieved cost optimization: ~$0/month vs $500+ for managed vector database alternatives
Enterprise Azure RAG System | GitHub: github.com/RicciJuaman/azure-enterprise-rag
Azure AI Search • Azure OpenAI • Document Intelligence • Python
- Architected production-grade RAG system for Queensland Education Department using Azure enterprise AI stack
- Configured document intelligence pipeline with automated cracking, chunking (512 tokens, 128 overlap), and
embedding generation through Azure AI Search skillsets
- Integrated Azure OpenAI for response generation with citation tracking and auditability
- Built for government sector with compliance-aware architecture and data sovereignty requirements
- System handles 1,320+ document chunks with hybrid ranking (vector + keyword + semantic)