Agentic Job Search

Home page
profile page
chatbot
Architecture diagram
sms flow
email flow

Inspiration

Job hunting is exhausting—students spend hours scrolling job boards while perfect opportunities slip by. We built an AI that does the heavy lifting: upload your resume once, get personalized job matches delivered daily.

What it does

Agentic Job Search uses multi-agent AI to automate your entire job search. It parses your resume with Amazon Nova Pro, matches you with relevant jobs through AWS Bedrock AgentCore, and delivers daily recommendations via email/SMS. Chat with the AI for career guidance—it remembers your goals across sessions.

How we built it

We architected a serverless, event-driven system with three components:

Multi-Agent AI (Orchestrator, Job Search, Career Advice agents) powered by Bedrock AgentCore
Real-time Pipeline: React → Lambda Proxy → AgentCore with streaming responses
Daily Automation: EventBridge triggers batch job matching (1 AM) and notifications (9 AM) via SQS/SES/SNS

Tech Stack: AWS Lambda, DynamoDB, S3 Vector Stores, Bedrock (Claude 4.5 Sonnet, Nova Pro, Titan Embeddings), React, TypeScript, CDK, Bedrock Agentcore, Strands Agent SDK

Challenges

Authentication: Bypassed Cognito session restrictions using Lambda Function URLs as proxy
Performance: Limited agents to 5 Knowledge Base queries max to reduce latency
Resume Parsing: Fine-tuned Nova Pro prompts with JSON schema validation for accuracy

What we learned

Multi-agent orchestration requires careful intent recognition. Serverless scales beautifully but demands robust error handling. AgentCore's memory transforms UX—personalization across sessions is game-changing.