Azure Copilot Plugin Development
Budget: $250 – $750 USD
I’m looking for a Microsoft-stack AI engineer who can turn our customer-support scenarios into production-ready Copilot plugins driven by Azure OpenAI, Azure Functions, and an agent-orchestration layer such as Semantic Kernel.
The natural-language work will run on Azure OpenAI (GPT-4 or GPT-3.5-Turbo). Business logic and data access belong in Azure Functions that you wire up through both Timer and Service Bus triggers; these functions will surface as OpenAI functions for the agent as well as REST endpoints for the plugin. The workflow itself needs to be orchestrated so the agent can decide which function to call and when—Semantic Kernel planners are preferred, but I’m open to an equivalent framework if you can show parity.
Primary use case: live customer support and interaction. The agent must automatically handle data entry and processing, updating customer records and returning a concise, context-aware answer back to the end user inside Copilot.
Deliverables
• Copilot plugin (OpenAPI manifest + description) registered and test-able inside Microsoft 365
• Azure Functions project with Timer & Service Bus triggered functions that call Azure OpenAI and our internal APIs
• Semantic Kernel (or equivalent) skills and planner code orchestrating the agentic flow
• Unit tests, sample prompts, and a README explaining setup, deployment, and extension points
Acceptance criteria
• End-to-end demo: user question in Copilot → agent selects correct function → data entered/updated → response returned in <5 s
• One-command build and deployment (Bicep/Terraform or ARM + GitHub Actions) with all tests passing
• Clear documentation enabling another engineer to reproduce the deployment from scratch
If you’ve already shipped similar Azure AI solutions and can blend OpenAI, Azure Functions (Timer & Service Bus), and agent frameworks, I’m eager to review your approach and timeline.
The natural-language work will run on Azure OpenAI (GPT-4 or GPT-3.5-Turbo). Business logic and data access belong in Azure Functions that you wire up through both Timer and Service Bus triggers; these functions will surface as OpenAI functions for the agent as well as REST endpoints for the plugin. The workflow itself needs to be orchestrated so the agent can decide which function to call and when—Semantic Kernel planners are preferred, but I’m open to an equivalent framework if you can show parity.
Primary use case: live customer support and interaction. The agent must automatically handle data entry and processing, updating customer records and returning a concise, context-aware answer back to the end user inside Copilot.
Deliverables
• Copilot plugin (OpenAPI manifest + description) registered and test-able inside Microsoft 365
• Azure Functions project with Timer & Service Bus triggered functions that call Azure OpenAI and our internal APIs
• Semantic Kernel (or equivalent) skills and planner code orchestrating the agentic flow
• Unit tests, sample prompts, and a README explaining setup, deployment, and extension points
Acceptance criteria
• End-to-end demo: user question in Copilot → agent selects correct function → data entered/updated → response returned in <5 s
• One-command build and deployment (Bicep/Terraform or ARM + GitHub Actions) with all tests passing
• Clear documentation enabling another engineer to reproduce the deployment from scratch
If you’ve already shipped similar Azure AI solutions and can blend OpenAI, Azure Functions (Timer & Service Bus), and agent frameworks, I’m eager to review your approach and timeline.