Application Architect

Automated Code Validation: Experience integrating DevSecOps tools (Checkov, tfsec, OPA) directly into automated pipelines to catch LLM hallucinations and control the “blast radius” before any code is applied. Zero-Trust & Air-Gapped Security: Must know how to build completely isolated network boundaries (no Internet Gateways) using AWS PrivateLink and VPC Endpoints to handle sensitive data securely. Cryptographic Domain Awareness: While they don’t need to be the primary Crypto SME, they must understand the baseline concepts they are architecting for-such as TLS 1.3 migrations, AES-256, Post-Quantum Cryptography (PQC), and parsing SBOM/CBOM JSON logs.

1. Agentic AI LangGraph Orchestration: Mandatory experience building cyclic state machines and multi-agent workflows using LangGraph (reject candidates who only have basic LangChain or chatbot experience). Native AWS Foundation Models: Deep experience with Amazon Bedrock and Anthropic Claude. They must know how to use strict system prompts and tool-calling to force the LLM to output valid infrastructure code. Advanced RAG Architecture: Experience designing Retrieval-Augmented Generation (RAG) systems with Amazon OpenSearch (or similar vector databases) to retrieve specific enterprise remediation rules and templates.

2. AWS ECS Fargate & Serverless Compute: Deep architectural experience running containerized workloads on AWS ECS Fargate, AWS Lambda, and designing event-driven systems. State Persistence (DynamoDB): Specific experience using DynamoDB for state management and checkpointing (crucial for pausing the AI while waiting for human approval). Asynchronous Integration: Mastery of decoupled AWS services like API Gateway, SQS, and EventBridge to handle long-running background tasks and Human-in-the-Loop (HITL) API callbacks.

3. Others Python Software Engineering: This is a hands-on architectural role. The candidate must be a strong Python developer capable of writing custom agentic logic, not just someone who configures dashboards. Terraform (IaC) Generation: Extensive knowledge of Terraform/HCL-specifically understanding how to instruct an AI to dynamically generate syntactically valid Terraform pull requests (Git PRs). Human-in-the-Loop (HITL) Design: Experience designing asynchronous workflows where an automated process pauses, waits for a human to review/approve via a UI, and gracefully resumes execution.