Senior Software Engineer - Agentic AI

Role

Senior Software Engineer focused on agentic AI: design, ship, and operate production-grade agentic systems that combine LLMs, tool/function calling, planning, memory, evaluation, and multi-agent communication. Work primarily in Python for AI services and integrate with the company’s enterprise stack to deliver reliable, cost-efficient, low-latency solutions in real customer workflows.

Key Responsibilities

• Build agentic AI applications on Azure AI Foundry using Azure OpenAI, Prompt Flow, tool/function-calling, evaluations, vector search, and orchestration for multi-step reasoning.
• Design and implement memory and grounding solutions (episodic/semantic/long-term) using vector/graph stores, and architect RAG pipelines and retrieval strategies to improve factuality and reduce latency/cost.
• Standardize tool/skill access via Model Context Protocol (MCP) and design agent-to-agent communication, delegation, and event-driven workflows.
• Connect agents to Microsoft data platforms (Microsoft Fabric, Dataverse, Databricks) ensuring lineage, governance, and auditability.
• Develop backend AI-native services in Python (FastAPI, asyncio) with evaluation harnesses, observability, and dashboards for cost/latency/quality.
• Embed AI features into TypeScript/Angular UIs and .NET/C# services, and integrate with SQL Server, NServiceBus, and Azure DevOps pipelines.
• Use AI-augmented development tools (GitHub Copilot, Bolt, Cursor, Replit) and vibe-coding workflows to accelerate delivery and improve code quality.
• Implement safety and reliability measures: guardrails, red-teaming, PII protection, prompt hardening, regression tests, automated evaluations, and uphold SLO/SLA standards.
• Drive full lifecycle agentic engineering including model/tool selection, API design, UI deployment monitoring, and continuous improvement.

Requirements

• Proven experience building LLM-powered applications with Azure OpenAI, embeddings, vector stores, RAG, prompt engineering, and evaluation pipelines.
• Hands-on experience with agent frameworks such as Semantic Kernel, LangGraph, LangChain Agents, AutoGen, or CrewAI.
• Ability to design deterministic, evaluatable, and safe agent behaviors including function schemas, tool success metrics, and fallback strategies.
• Practical experience with Prompt Flow for authoring, testing, and deploying multi-step AI workflows.
• Experience building and consuming MCP services, memory architectures (episodic, semantic, vector, graph), and agent-to-agent communication patterns.
• Integration experience with Microsoft Fabric, Databricks, Dataverse, Supabase, and related data platforms for grounding and telemetry.
• Familiarity with Azure services (App Service/Functions/AKS, Key Vault, Storage, Event Hubs/Service Bus, Monitor/Application Insights) and Azure IoT Hub/IoT Edge for device telemetry scenarios.
• Production-grade Python skills (FastAPI, asyncio, type hints), Postgres/SQL, Redis, queues, OpenTelemetry, CI/CD, containerization, API design, and testing practices.
• Experience with TypeScript/Angular for operator consoles and integration with .NET/C#, SQL Server, NServiceBus, and Azure DevOps.
• Strong mentoring, communication, ownership, and on-call reliability practices.

Tech Stack (explicit)

Azure AI Foundry, Azure OpenAI, Prompt Flow, Azure AI/Cognitive Search, MCP, Semantic Kernel, LangGraph, LangChain, AutoGen, CrewAI, HuggingFace embeddings, vector DBs, RAG, Databricks, Microsoft Fabric (OneLake/Lakehouse/Warehouse/Real-Time), Dataverse, Supabase, Azure IoT Hub, IoT Edge, App Service, Azure Functions, AKS, Key Vault, Storage, Event Hubs, Service Bus, Monitor/Application Insights, Python (FastAPI, asyncio), Postgres, SQL Server, Redis, OpenTelemetry, NServiceBus, .NET/C#, TypeScript/Angular, Ionic, Cypress, GitHub Copilot, Bolt, Cursor, Replit, MLflow, Delta Lake.

Location & Work Model

Hybrid role based in Dunwoody.

Outcomes / Success Measures

• Improved agent tool-use success rates and reduced hallucination/retry rates.
• Meeting latency (P50/P95) and token-cost budgets with measurable efficiency gains.
• Shipped features adopted by real users and demonstrable KPI improvements.
• High test coverage, stable CI/CD, observable systems, and healthy on-call posture.