Artificial Intelligence Engineer

About the role

About the role: Own the design, development, deployment, and quality of AI agents on AWS using Amazon Bedrock foundation models and Amazon Bedrock AgentCore. This hands-on engineering role requires direct experience building and deploying agents—not solely prompt design or the use of AI coding assistants. Work with client developers to establish a reusable reference implementation, deliver a five-template agent pilot, and transfer secure, testable engineering patterns that the client can sustain. What you will do at Sogeti: Design and implement agents using Amazon Bedrock models and deploy them through AgentCore Runtime, using Strands Agents, LangGraph, or an appropriate custom framework. Configure each agent archetype's instructions, context, tool definitions, orchestration, session handling, and structured result contracts; implement AgentCore Memory where persistent context is required. Connect agents to enterprise APIs, AWS Lambda functions, and MCP tools using AgentCore Gateway and Model Context Protocol integrations as appropriate. Implement least-privilege IAM permissions, secure credential handling, and agent authentication and authorization, using AgentCore Identity where appropriate. Build the first golden agent template as the reference implementation for the other four, including configuration, deployment automation, documentation, and reusable integration patterns. Build output verification, path and tool allowlists, schema validation, API contract tests, automated agent evaluations, and review-ready pull request structures. Define human approval points for sensitive actions. Instrument agents using AgentCore Observability and Amazon CloudWatch to troubleshoot execution, tool calls, errors, latency, and token consumption. Run the pilot tuning loop: analyze failures, evaluation results, and reviewer feedback; improve prompts, context, tools, templates, and model configuration against agreed acceptance criteria. Automate repeatable AWS deployments with infrastructure as code and CI/CD; document environment configuration, operational procedures, and rollback approaches. Pair with the client's API developer on templates 2 through 5 and transfer architecture, testing, troubleshooting, and operational knowledge. Success Criteria: A working golden agent template deployed to the agreed AWS environment using AgentCore Runtime and Amazon Bedrock models. Five reusable agent templates delivered with the client team, with documented configurations, tool interfaces, result contracts, and deployment instructions. An automated verification and evaluation suite, with acceptance criteria and baseline measures for task success, output quality, latency, and token consumption agreed with the client. Documented security controls, operational monitoring, troubleshooting guidance, and knowledge transfer enabling client developers to extend and maintain the agents. What you will bring (Required): 5+ years building production APIs on AWS, including REST and at least one of GraphQL or event-driven architectures. Demonstrated hands-on experience developing and deploying AI agents using Amazon Bedrock models and Amazon Bedrock AgentCore, including direct implementation experience with AgentCore Runtime. Hands-on experience with at least one additional AgentCore capability: Gateway, Memory, Identity, or Observability. General familiarity with Bedrock or use of coding assistants alone does not satisfy the AgentCore requirement. Strong Python and/or TypeScript development skills, AWS SDK experience, and experience implementing agent orchestration and tool calling with an agent framework or custom code. Experience integrating agents with enterprise APIs and tools, including MCP-based integrations or AgentCore Gateway. Practical AWS engineering skills with IAM, Lambda, API Gateway, CloudWatch, and services such as S3 or DynamoDB; ability to implement secure access and troubleshoot distributed applications. Strong testing discipline: OpenAPI, unit and integration testing, contract tests, structured-output validation, and automated evaluation of agent behavior and generated code. Experience with Git, pull request workflows, CI/CD, and infrastructure as code using AWS CDK, AWS CloudFormation, or Terraform. Ability to explain agent architecture, security controls, behavior, and failure modes clearly, and to pair effectively with client engineering teams. What you will bring (Preferred): Experience operating AgentCore-based agents in production, including incident troubleshooting, performance tuning, and cost optimization. Experience across multiple AgentCore capabilities, including Memory, Identity, Gateway, and Observability. Experience with Amazon Bedrock Knowledge Bases, retrieval-augmented generation, and Amazon Bedrock Guardrails. Experience building MCP servers, multi-agent workflows, or agent-to-agent integrations. Built reusable templates, scaffolding, or code generators adopted by other engineering teams; measured acceptance rates and compared agent configurations. Hands-on use of Kiro, Claude Code, or similar coding agents in specification-driven or headless workflows beyond IDE autocomplete. Experience with BMAD or similar specification-driven development methods. Life at Sogeti: Sogeti supports all aspects of your well-being throughout the changing stages of your life and career. For eligible employees, we offer: Flexible work options 401(k) with 150% match up to 6% Employee Share Ownership Plan Medical, Prescription, Dental

What you'll do

• Own the design, development, deployment, and quality of AI agents on AWS using
• Amazon Bedrock foundation models and Amazon Bedrock AgentCore.
• This hands-on engineering role requires direct experience building and deploying agents—not solely prompt design or the use of AI coding assistants.
• Work with client developers to establish a reusable reference implementation, deliver a five-template agent pilot, and transfer secure, testable engineering patterns that the client can sustain.
• at Sogeti: Design and implement agents using
• Amazon Bedrock models and deploy them through AgentCore Runtime, using
• Strands Agents, LangGraph, or an appropriate custom framework.
• Configure each agent archetype's instructions, context, tool definitions, orchestration, session handling, and structured result contracts; implement
• AgentCore Memory where persistent context is required.
• Connect agents to enterprise APIs, AWS Lambda functions, and MCP tools using
• AgentCore Gateway and Model Context Protocol integrations as appropriate.
• Implement least-privilege IAM permissions, secure credential handling, and agent authentication and authorization, using
• AgentCore Identity where appropriate.
• Build the first golden agent template as the reference implementation for the other four, including configuration, deployment automation, documentation, and reusable integration patterns.
• Build output verification, path and tool allowlists, schema validation, API contract tests, automated agent evaluations, and review-ready pull request structures.
• Define human approval points for sensitive actions.
• Instrument agents using
• AgentCore Observability and Amazon CloudWatch to troubleshoot execution, tool calls, errors, latency, and token consumption.
• Run the pilot tuning loop: analyze failures, evaluation results, and reviewer feedback; improve prompts, context, tools, templates, and model configuration against agreed acceptance criteria.
• Automate repeatable AWS deployments with infrastructure as code and CI/CD; document environment configuration, operational procedures, and rollback approaches.
• Pair with the client's API developer on templates 2 through 5 and transfer architecture, testing, troubleshooting, and operational knowledge.
• Success Criteria: A working golden agent template deployed to the agreed AWS environment using
• AgentCore Runtime and Amazon Bedrock models.
• Five reusable agent templates delivered with the client team, with documented configurations, tool interfaces, result contracts, and deployment instructions.
• An automated verification and evaluation suite, with acceptance criteria and baseline measures for task success, output quality, latency, and token consumption agreed with the client.
• Documented security controls, operational monitoring, troubleshooting guidance, and knowledge transfer enabling client developers to extend and maintain the agents.

What you'll bring

• (Required): 5+ years building production APIs on AWS, including REST and at least one of GraphQL or event-driven architectures.
• Demonstrated hands-on experience developing and deploying AI agents using
• Amazon Bedrock models and Amazon Bedrock AgentCore, including direct implementation experience with AgentCore Runtime.
• Hands-on experience with at least one additional
• AgentCore capability: Gateway, Memory, Identity, or Observability.
• General familiarity with Bedrock or use of coding assistants alone does not satisfy the AgentCore requirement.
• Strong Python and/or
• TypeScript development skills, AWS SDK experience, and experience implementing agent orchestration and tool calling with an agent framework or custom code.
• Experience integrating agents with enterprise APIs and tools, including MCP-based integrations or AgentCore Gateway.
• Practical AWS engineering skills with IAM, Lambda, API Gateway, CloudWatch, and services such as S3 or DynamoDB; ability to implement secure access and troubleshoot distributed applications.
• Strong testing discipline: OpenAPI, unit and integration testing, contract tests, structured-output validation, and automated evaluation of agent behavior and generated code.
• Experience with Git, pull request workflows, CI/CD, and infrastructure as code using AWS CDK, AWS CloudFormation, or Terraform.
• Ability to explain agent architecture, security controls, behavior, and failure modes clearly, and to pair effectively with client engineering teams.
• (Preferred): Experience operating
• AgentCore-based agents in production, including incident troubleshooting, performance tuning, and cost optimization.
• AgentCore capabilities, including Memory, Identity, Gateway, and Observability.
• Experience with Amazon Bedrock Knowledge Bases, retrieval-augmented generation, and Amazon Bedrock Guardrails.
• Experience building MCP servers, multi-agent workflows, or agent-to-agent integrations.
• Built reusable templates, scaffolding, or code generators adopted by other engineering teams; measured acceptance rates and compared agent configurations.
• Hands-on use of Kiro, Claude Code, or similar coding agents in specification-driven or headless workflows beyond IDE autocomplete.
• Experience with BMAD or similar specification-driven development methods.
• Life at Sogeti: Sogeti supports all aspects of your well-being throughout the changing stages of your life and career.
• For eligible employees, we offer: Flexible work options 401(k) with 150% match up to 6%
• Employee Share Ownership Plan Medical, Prescription, Dental