ricci-juaman
Ricci Juaman | AI Engineer
| **Production AI Systems | RAG Architectures | Full-Stack Development** |
📍 Brisbane, Queensland, Australia
📧 juamandudz1590@gmail.com
💼 LinkedIn
🔓 Open to opportunities | Available immediately
👋 About Me
I’m an AI Engineer specializing in production RAG (Retrieval-Augmented Generation) systems with full-stack capabilities. I build end-to-end AI applications from architecture design through deployment, with experience in both custom implementations and enterprise-scale solutions.
What makes me different:
- 🔧 Built RAG systems from scratch AND with enterprise tools (understand both approaches)
- 🏗️ Full-stack: Backend (Python, FastAPI, PostgreSQL) + Frontend (React) + DevOps (CI/CD, Azure)
- 💰 Cost-conscious: Evaluate build vs buy tradeoffs based on business requirements
- 🤝 Community: Active in Brisbane AI community, mentor software engineers transitioning to AI/ML
- 📊 Business-minded: 5+ years solving real-world problems in pricing strategy, analytics, and data
🚀 Featured Projects
1. Custom Full-Stack RAG System with Hybrid Search
Production-grade retrieval system built from first principles
Tech Stack: Python • React • FAISS • PostgreSQL • FastAPI • BGE-Large • BM25 • Azure • CI/CD
What I Built:
- Custom Vector Database: Built from scratch using FAISS with BGE-Large embeddings
- Hybrid Search Engine: Combined semantic similarity (vector search) with lexical matching (BM25) for superior retrieval accuracy
- Full-Stack Application: React frontend with real-time search + FastAPI backend + PostgreSQL metadata layer
- Production Infrastructure: CI/CD pipelines, automated testing, monitoring, deployment on Azure
Architecture:
Key Technical Decisions:
Why build custom instead of using Pinecone/Weaviate?
- Cost optimization: ~$0/month vs $500+/month for managed services
- Learning objective: Deep understanding of vector search, embedding models, retrieval algorithms
- Control: Fine-grained tuning of hybrid search weights and retrieval strategies
Why hybrid search (semantic + lexical)?
- Pure semantic search fails on exact matches (names, codes, acronyms)
- Pure keyword search misses conceptual queries
- Hybrid approach handles both: “What is inclusive teaching?” (semantic) AND “Policy code 234B” (lexical)
Performance:
- 📄 50000 documents indexed
- ⚡ <4s average query latency - (Can be improved)
- 🎯 ~85% retrieval accuracy (measured on test queries)
- 💾 Efficient memory usage with FAISS IVF indexing
What This Demonstrates:
✅ Understanding of embedding models and vector search fundamentals
✅ Algorithm implementation (BM25, score fusion, ranking)
✅ Full-stack development (frontend, backend, database)
✅ Production engineering (testing, CI/CD, deployment)
✅ Cost-benefit analysis and architectural decision-making
2. Enterprise Azure RAG System
Production AI solution for Queensland Education Department
Tech Stack: Azure AI Search • Azure OpenAI • Document Intelligence • Blob Storage • Python • Skillsets
What I Built:
- Enterprise RAG Architecture: Production-ready system using Azure’s managed AI services
- Document Intelligence Pipeline: Automated document cracking, parsing complex formats (PDF, HTML, DOCX)
- Smart Chunking Strategy: Skillset-based chunking with overlap for context preservation
- Compliance-Aware: Built for government sector with data sovereignty and audit requirements
- Automated Workflows: End-to-end pipeline from document upload to searchable knowledge base
System Flow:
Documents Upload (Blob Storage)
↓
Document Intelligence
(Crack, Parse, OCR)
↓
Azure AI Search Skillset
(Chunk, Embed, Enrich)
↓
Hybrid Index
(Vector + Keyword + Semantic)
↓
Azure OpenAI Integration
↓
Responses with Citations
Key Technical Decisions:
Why Azure managed services instead of custom?
- Enterprise requirements: Compliance, SLA guarantees, data sovereignty for government sector
- Time to market: Document Intelligence handles complex parsing (PDFs, forms) that would take weeks to build
- Scalability: Managed infrastructure handles load spikes automatically
- Maintenance: Azure manages updates, security patches, infrastructure
Chunking Strategy:
- Fixed-size chunking (current): 512 tokens with 128 token overlap
- Semantic chunking (planned): Chunk by document structure (sections, subsections) for legal documents
- Trade-offs evaluated: Speed vs accuracy, context preservation vs index size
Performance:
- 📄 7 HTML files → 1,320 chunks (demonstrates pipeline scalability)
- ⏱️ ~3 seconds query latency on free tier (sub-second on paid tier)
- 🔍 Hybrid ranking: Vector + keyword + semantic reranking
- 📚 Ready to scale to 100K+ documents
What This Demonstrates:
✅ Enterprise architecture and managed service expertise
✅ Azure AI stack proficiency (AI Search, OpenAI, Document Intelligence)
✅ Document processing and ETL pipeline design
✅ Compliance and governance awareness
✅ Evaluation of build vs buy tradeoffs
3. Legal RAG System (In Development)
Domain-specific RAG with document structure chunking
Tech Stack: Python • Azure AI Search • Custom Chunking Logic • Queensland Legislation
Planned Features:
- Hierarchical Chunking: Preserve legal document structure (Act → Part → Division → Section)
- Citation Accuracy: Critical for legal applications - track exact sources
- Cross-Reference Linking: Connect related statutes and precedents
- Jurisdiction Awareness: Handle Queensland vs Federal legislation distinctions
Why Legal RAG is Complex:
Legal documents require specialized handling that generic RAG can’t provide:
- 📋 Structure matters: A section means nothing without its parent Act context
- 🎯 Citation precision: Lawyers need exact sources, not “approximately from X”
- 🔗 Relationships: Statutes reference each other, amendments supersede originals
- ⚖️ Jurisdiction: Same topic, different rules by state/federal level
Target Timeline: Q1 2026
Goal: Portfolio piece demonstrating domain specialization
🛠️ Technical Skills
AI/ML
- RAG Systems: Architecture, implementation, evaluation, optimization
- Vector Databases: FAISS (custom implementation), Azure AI Search, Pinecone (familiar)
- Embeddings: BGE-Large, E5, OpenAI, Cohere (model selection and benchmarking)
- Search Algorithms: Hybrid search, BM25, semantic search, reranking strategies
- LLMs: Azure OpenAI, prompt engineering, response generation, citation handling
Backend & Data
- Languages: Python (production-level), SQL, JavaScript
- Frameworks: FastAPI, Flask, LangChain, LlamaIndex
- Databases: PostgreSQL (schema design, optimization), SQL querying, indexing strategies
- APIs: RESTful API design, endpoint development, documentation
Frontend
- Frameworks: React, JavaScript, HTML/CSS
- UI/UX: Real-time search interfaces, data visualization, responsive design
Cloud & DevOps
- Azure: AI Search, OpenAI, Blob Storage, Document Intelligence, deployment
- CI/CD: GitHub Actions, automated testing, deployment pipelines
- Tools: Docker, Git/GitHub, pytest, monitoring/logging
Data & Analytics
- Experience: 5+ years in data analytics, pricing strategy, business intelligence
- Skills: Statistical modeling, A/B testing, ETL pipelines, data visualization
- Tools: Python (pandas, numpy), SQL, Salesforce, Power BI
💼 Professional Experience Highlights
Shiply | Pricing Strategy Manager
March 2024 - Present | Remote (UK)
Applied data-driven approaches to pricing optimization:
- Design and execute A/B tests for pricing experiments
- Build analytical pipelines and predictive models for price elasticity
- Collaborate with engineering on API integration and data infrastructure
- Translate complex analysis into executive-level insights
Relevance to AI Engineering: Statistical testing, model evaluation, API integration, stakeholder communication
Acrewell | Data Analyst
July 2022 - November 2023 | Philippines
Property valuation and analytics for land acquisition:
- Multi-criteria evaluation (accessibility, topography, utilities, soil type)
- Comparative analysis for pricing precision
- Built reports and dashboards in Salesforce
- Proactive research with county offices for due diligence
Relevance to AI Engineering: Multi-modal data evaluation, risk assessment, system integration
Radarr (Genesys) | Data Analyst
March 2022 - June 2022 | Singapore
Social listening platform - ETL, analysis, and visualization:
- Full data lifecycle management
- Pattern recognition in noisy data (signal vs noise)
- Cross-functional collaboration with product teams
Relevance to AI Engineering: Data pipeline design, noise handling (critical for RAG retrieval quality)
🎓 Education & Continuous Learning
Bachelor of Arts - Philosophy | Notre Dame of Marbel University
2008 - 2012
Philosophy training provides:
- First-principles thinking and problem decomposition
- Logical reasoning and argument construction
- Abstract pattern recognition across domains
- Meta-learning skills (learning how to learn)
Google Professional Data Analytics Certificate | Coursera
2021 - 2022
Self-Directed AI/ML Learning:
- RAG systems, vector databases, embeddings, LLMs
- Full-stack development (React, FastAPI, PostgreSQL)
- Azure AI services, DevOps, CI/CD pipelines
- Built production systems while learning (learning by doing)
🤝 Community & Collaboration
Brisbane AI Community
- Mentorship: Guide software engineers transitioning into AI/ML roles
- Collaboration: Currently working with QUT software engineers on production AI projects
- Knowledge Sharing: Regular discussions on RAG architectures, vector databases, retrieval optimization
- Engagement: 4+ engineers have reached out for advice and guidance on AI career transitions
Why This Matters:
- 💡 Teaching solidifies understanding (Feynman technique)
- 🌐 Active local network (Brisbane tech community)
- 🤝 Demonstrates collaboration and communication skills
- 📈 Others seek my expertise (validation of knowledge)
📝 Technical Writing & Content
Planned Articles (Q1 2026):
- “Building a Production RAG System: Custom vs Enterprise” - Architecture decisions, cost analysis, tradeoffs
- “Hybrid Search Explained: When Semantic Search Isn’t Enough” - Technical deep-dive with benchmarks
- “Legal RAG: Why Document Structure Matters” - Domain-specific RAG challenges
Why Writing Matters:
- Demonstrates communication skills (critical for senior roles)
- Solidifies technical understanding through teaching
- Builds personal brand and visibility
- Contributes to broader AI engineering community
🎯 What I’m Looking For
Role: AI Engineer or ML Engineer (mid to senior level)
Location: Brisbane preferred, open to Sydney/Melbourne, remote considered
Type: Full-time, contract, or contract-to-hire
Ideal Environment:
- 🚀 Building production AI products (not just research/POCs)
- 🏗️ Architecture decisions and technical ownership
- 🤝 Collaborative team that values knowledge sharing
- 📈 Growth opportunities and technical challenges
- 🎓 Learning culture and mentorship (both ways)
What I Bring:
- Production RAG experience (custom + enterprise)
- Full-stack capabilities (ship complete features)
- Cost-conscious engineering (build vs buy thinking)
- Business context (5+ years solving real problems)
- Mentorship and collaboration skills
- Self-directed learning and adaptability
📬 Let’s Connect
I’m actively seeking opportunities and happy to discuss:
- RAG system architectures and implementation challenges
- Vector database tradeoffs (custom vs managed)
- AI engineering career transitions and mentorship
- Brisbane AI community events and collaboration
Fastest response:
- 📧 Email: juamandudz1590@gmail.com
- 💼 LinkedIn: linkedin.com/in/ricci-doddz-juaman
- 📱 Phone: +61 494 376 293
Portfolio & Projects:
- 🔗 This Portfolio:
ricci-juaman.github.io/portfolio(you’re here!) - 💻 GitHub: [github.com/RicciJuaman]
🏆 Quick Stats
📊 Production RAG Systems Built: 2 (custom + enterprise)
💻 Lines of Code (estimated): 15,000+
🏗️ Full-Stack Features Shipped: Multiple (frontend + backend + AI)
👥 Engineers Mentored: 4+
📈 Years in Data/Analytics: 5+
🎓 Self-Directed Learning Projects: 10+ (RAG, vector DBs, full-stack, cloud)
☁️ Cloud Platforms: Azure (production), AWS (learning)
📌 Pinned Repositories
custom-rag-system
Full-stack RAG with FAISS, hybrid search, React UI, PostgreSQL. Production-grade retrieval system built from scratch.
azure-enterprise-rag
Enterprise RAG using Azure AI Search, Document Intelligence, OpenAI. Built for Queensland Education Department.
legal-rag-prototype
(In Development) Domain-specific RAG for legal documents with structure-aware chunking.
Last Updated: December 28, 2024
⭐ If you find my work interesting, let’s connect! I’m always happy to discuss RAG architectures, AI engineering, or grab coffee in Brisbane.