Core Competencies: Agentic AI Systems | RAG Architecture | Model Fine-Tuning | Real-Time AI Applications | LLM Orchestration

Recent Projects (2024–Present)

1. Experimental Agent with Hermes

Built an experimental agent on Hermes, running qwen3.6:35b-a3b locally on an AMD machine with 64GB RAM. All workflow orchestration and domain expertise are modularized into dedicated skills.

Demo: https://x.com/shawn810720/status/2060224726900121767

2. AI Product Consulting Agent (MVP)

Architected an intelligent routing system for product consulting.

• Intent classification: Fine-tuned Gemma3:270m for semantic intent routing

• Model pipeline: Multi-model Ollama setup (Gemma3:4b, Nomic-Embed-Text v1.5)

• Retrieval: Advanced RAG with Small-to-Big retrieval and reranking

• Knowledge store: Qdrant vector database for product knowledge

• Interface: Real-time SSE streaming UI

• Tech stack: LangGraph, LlamaIndex, Flask

Demo: https://x.com/shawn810720/status/1975782475910816049

3. Knowledge Base Q&A System

Production-grade product Q&A platform using Llama 3 and DeepSeek R1 for accurate information retrieval.

Demo: https://x.com/shawn810720/status/1910602758933405908

4. Real-Time Vision AI Application

Live multimodal processing pipeline powered by Llama 3.2-Vision.

Demo: https://x.com/shawn810720/status/1897959157166428644

5. DayOff System: Chain-of-Thought Framework

Engineered a reasoning framework that gives Llama 3 8B System 2 thinking capabilities through structured workflow optimization.

Demo: https://x.com/shawn810720/status/1892398443957518769