Godfrey Yang

About Me

3 min read

Godfrey Yang

AI Agent Engineer focused on LLM systems, RAG pipelines, backend APIs, and secure automation.

I build practical AI systems that connect models to tools, data, and real operational workflows. My work sits between software engineering, data engineering, DevOps, and cybersecurity: designing agent workflows, integrating APIs, building retrieval pipelines, automating data quality checks, and keeping systems auditable enough for serious environments.

What I Build

  • AI agents and tool-using systems using MCP, Agent Skills, tool use, ACI design, orchestrator-worker patterns, and evaluator-optimizer loops.
  • RAG and retrieval systems with Qdrant, PostgreSQL, Ollama, reranking, Cog-RAG, LightRAG, and Neo4j-style knowledge graph exploration.
  • Backend and data automation across REST APIs, Python, TypeScript, SQL, Snowflake, Power BI, Azure DevOps, Docker, and CI/CD pipelines.
  • Security-aware AI workflows grounded in IAM/PAM, SIEM, zero trust, auditability, guardrails, OWASP, Essential 8, and ISO 27001.

Current Focus

I am currently exploring how to make AI agents more useful in production-style workflows: better context engineering, safer tool use, repeatable evaluation, local-first inference, and cost-aware model selection.

Recent work includes benchmarking coding agents, comparing local LLM inference stacks, building private RAG document pipelines, and documenting what actually works when models need to retrieve information, call tools, and complete multi-step tasks.

Professional Background

I have 6+ years of experience delivering engineering, automation, data, and security solutions across consulting, energy, resources, and government environments.

At Deloitte, I work across AI, data, and software engineering projects, including API-driven reporting automation, ML-ready data pipelines, ETL/data quality workflows, and DevOps tooling for enterprise clients. Earlier roles gave me hands-on experience in security operations, privileged access management, SIEM integration, distributed infrastructure, and systems automation.

My formal background is in Mechatronics and Automation Engineering, with additional training in Cyber Security. That mix shapes how I approach AI systems: useful automation first, strong engineering foundations, and clear attention to failure modes, controls, and operational reliability.

Selected Work

  • Built a local RAG pipeline achieving 93% retrieval accuracy across multiple modes using n8n, Ollama, Qdrant, PostgreSQL, reranking, Cog-RAG, and LightRAG/Neo4j.
  • Benchmarked 10+ LLMs and coding agents across OpenAI, Anthropic, Google, Ollama, llama.cpp, vLLM, and AMD RX 7800 XT local inference.
  • Automated 1,000+ monthly enterprise data workflows, reducing reporting turnaround from 2 days to 2 hours.
  • Built Python/R ETL and QA pipelines improving data accuracy from 75% to 95% and reducing processing time by 60%.
  • Deployed CyberArk EPM and Wazuh SIEM/IDS workflows for zero-trust endpoint management, privileged access control, and threat monitoring.

Why This Site Exists

This site is my technical notebook for AI engineering, LLM systems, RAG, agents, local inference, cybersecurity, and automation. I use it to publish experiments, implementation notes, benchmark results, and lessons learned from building systems rather than only reading about them.

If you are hiring for AI agent engineering, LLM application development, backend automation, RAG systems, or security-aware AI tooling, the best starting points are my project writeups and GitHub work.

Contact

Graph View