Software Engineer · beginner
Conversational AI Agent with Tool Use
An AI agent that doesn't just chat — it actually does things. Calls APIs, queries databases, triggers workflows. Built on Amazon Bedrock AgentCore.
Project overview
Chatbots are table stakes. Agents that can actually DO things — call APIs, query databases, trigger workflows — are what companies are racing to build right now. Showing you understand agentic patterns puts you ahead of 90% of applicants.
Real-world use case
A personal assistant that checks your calendar, adds events, searches the web, and summarizes emails. Or a customer service agent that looks up orders and processes returns.
Tools and technologies
- Amazon Bedrock AgentCore
- AWS Lambda
- DynamoDB
- API Gateway
Skills you can demonstrate
Function Calling, Tool Definitions, Cloud Integration
Build roadmap
- Set up AWS + Bedrock access
Create an AWS account. Request access to Claude or Nova in Bedrock — approval can take a few hours. Install the AWS CLI. - Define the agent's tools
Pick 2–3 tools (e.g., get_calendar, add_event, search_web). Write JSON schemas describing each tool's inputs and outputs. - Build a Lambda for each tool
Each tool is a Python Lambda that does one thing well. Test each independently before wiring to the agent. - Configure AgentCore
Create the agent in AgentCore. Attach the Lambdas as tools. Write the system prompt that tells the agent when to call each one. - Test multi-step flows
Try queries that require 2+ tool calls in sequence ("What's on my calendar tomorrow, and add a 30-minute prep block before each meeting"). Iterate the prompt until it works. - Wrap with API Gateway
Expose the agent via a public REST endpoint. Build a tiny web UI so recruiters can talk to your agent live. That demo is your portfolio.
Ways to extend the project
Add memory so the agent remembers prior turns. Layer in multi-agent collaboration where specialized agents hand off tasks to each other.
How to showcase your work
- GitHub: Detailed README with the AgentCore config, Lambda code, and a demo video of the agent calling tools end-to-end.
- LinkedIn: Post a video of the agent answering a multi-step question that requires 2+ tool calls. Caption: "Not a chatbot. An agent."
- Resume: Focus on "Function calling", "Tool definitions", and "Production AI agents on AWS".