AI DevOps & Infrastructure
AI Agent
Inline: Ground-up rebuild of inference services, agentic solution, and custom MCP
"Tech 42 was an exceptional partner throughout our production agent build and launch. Over the course of the engagement, their team helped us move to a production-ready MCP-driven agentic architecture on AWS, rebuilding our inference service from the ground up to support scale, performance, and enterprise readiness.
Together, we implemented a robust agent architecture using AWS Bedrock and LangGraph, introduced dynamic guardrails and system prompts per publisher, integrated secure MCP tooling with OAuth, and added critical capabilities around observability, caching, and cost-aware execution.
Beyond the technical execution, Tech 42 operated as a true extension of our team. Communication was proactive and highly responsive, and the team consistently demonstrated ownership, diligence, and flexibility as we worked through real-world constraints and last-mile challenges. Their work directly enabled us to scale our platform, unlock additional AWS credits, and confidently onboard enterprise customers.
We are extremely pleased with the outcomes of this engagement and grateful for the partnership. We would gladly work with Tech 42 again and recommend them to any organization looking to productionize advanced AI systems on AWS."
Sam Doliner
Co-founder
Project summary
Inline needed to move a proof-of-concept LLM agent into production. The original POC was functional, but the team was ready to productionize with guardrails, authentication, caching, and infrastructure required for enterprise deployment. Tech 42 rebuilt the agent architecture from the ground up using Amazon Bedrock, SageMaker, and ECS. The production system introduced dynamic guardrails and system prompts configurable per publisher, integrated secure OAuth tooling, added observability and query caching with ElastiCache, and deployed as Infrastructure as Code using CDK. The result is a scalable, cost-efficient agentic architecture capable of powering conversational engagement across multiple publishers. Inline can now onboard enterprise customers with confidence and operate the platform at scale.
