# 🧠 Building Effective AI Agents on AWS

Agentic AI is a game-changer. Instead of relying on a single massive model for every task, we orchestrate specialized tools, models, and workflows in modular, scalable ways. These agentic systems allow us to reason, decide, and act in cost-efficient and context-aware pipelines.

Why agents instead of plain LLMs?

• 🧠 Smarter: Tools + models = decision-making with memory and structure
• 💸 Cheaper: Use smaller models/tools only when needed (pay-as-you-go)
• 🧰 Modular: Add/replace tools without retraining everything
• 📊 Traceable: With observability tools like Langfuse
• ⚡ Faster to Iterate: Custom logic with better control

---

🌟 This Repo Is Based On:

Hands-on work from the AWS GenAI Agents Workshop, focusing on building, fine-tuning, and deploying agentic AI systems using:

• Amazon Bedrock
• Amazon SageMaker
• Strand
• CrewAI
• LangGraph
• Langfuse for observability

⚠️ This is a refined, simplified version of the original AWS workshop repo, meant for reproducibility and educational sharing.

---

📌 Workshop Highlights

• ✅ Fine-tuned and deployed an LLM using SageMaker Studio
• ✅ Used Amazon Bedrock to access foundation models like Claude, LLaMA 3
• ✅ Defined tools for DSL (OpenSearch, GuardDuty) and blog summarization
• ✅ Built agents using CrewAI and Strands
• ✅ Chained agents using Evaluator-Optimizer and Orchestrator-Worker patterns
• ✅ Enabled Langfuse observability to monitor agent behavior

---

🗂 Repo Sections

Folder	Purpose
0-setup-environment/	Setup virtual environment and SageMaker Studio
1-agent-inference/	Use Bedrock and SageMaker for inference
2-tools-and-integration/	Build and register custom tools
3-building-agents/	Define simple and complex agents
4-frameworks/	Examples using CrewAI and Strands
5-observability/	Langfuse integration for API tracing
special-usecases/	Text-to-SQL, DSL with MCP server, and more

---

🔧 Setup Instructions

git clone https://github.com/YOUR_USERNAME/aws-agentic-ai-workshop.git
cd aws-agentic-ai-workshop
python -m venv genai-env
source genai-env/bin/activate      # or use genai-env\Scripts\activate on Windows
pip install -r requirements.txt
jupyter lab

---

📣 Credits

Created during AWS GenAI Workshop
Curated & restructured by Sahar Zargarzadeh