Serverless Infrastructure Quality Engineer

Obsidian; San Francisco (CA); Remote; USD 120,000 - 150,000; Full time

Job summary

Obsidian is seeking an experienced reviewer to evaluate the quality and correctness of AWS serverless and IaC tasks used in training and evaluating frontier AI models. The role focuses on multi-service serverless designs, IaC fidelity, and cross-service integration, with rubric-based feedback to guide improvements.

The candidate will examine event-driven architecture, IAM boundaries, and failure/retry semantics, including comparisons between LocalStack simulations and real AWS behavior.

Qualifications

• 3+ years hands-on experience building multi-service AWS serverless applications.
• Strong experience with infrastructure-as-code (CDK/CloudFormation/SAM).
• Demonstrated ability to assess event-driven architecture correctness, IAM/permission boundaries, and failure/retry semantics.
• Familiarity with local emulation/testing fidelity (e.g., LocalStack) versus real AWS behavior.

Responsibilities

• Evaluate the quality, correctness, and cloud-architecture soundness of AWS serverless tasks used to train and evaluate frontier AI models.
• Provide rubric-based written feedback detailing strengths, gaps, and concrete improvement steps.
• Assess cross-service integration correctness and IAM/permission boundaries across services.
• Document findings in a structured report with actionable recommendations.

Skills

AWS serverless

IaC

Event-driven design

IAM boundaries

LocalStack testing

Tools

AWS CDK

CloudFormation

SAM

LocalStack

Lambda

API Gateway

DynamoDB

S3

Step Functions

EventBridge

Job description

Obsidian is seeking an experienced reviewer to evaluate the quality and correctness of AWS serverless and IaC tasks used in training and evaluating frontier AI models. The role focuses on multi-service serverless designs, IaC fidelity, and cross-service integration, with rubric-based feedback to guide improvements.

The candidate will examine event-driven architecture, IAM boundaries, and failure/retry semantics, including comparisons between LocalStack simulations and real AWS behavior.