01Overview
Overview The AI Observability Engineer (Agentic Frameworks & AI Agent Operations Center Developer) builds and operationalizes agentic AI solutions using modern orchestration frameworks and contributes to an AI Agent Operations Center that enables safe, reliable, and observable agent behavior at scale. This role focuses on developing agent workflows (planning, tool execution, memory, and RAG), integrating guardrails and evaluations, and delivering operational capabilities such as run management, telemetry, and incident triage for production agents. Responsibilities
AI Agent Operations Center (70%) Build operations center capabilities for agent runtime management: agent registry, versioning, deployment tracking, and run histories
Enable operational workflows such as incident triage, replay/debug runs, trace correlation, and root-cause analysis across agent steps
Implement operational dashboards and views for agent health: success rate, latency, tool failure rate, cost per run, and loop detection Instrument agent flows end-to-end using OpenTelemetry (or equivalent), enabling correlation across prompts, tool calls, retrieval, and responses
Implement semantic conventions and tagging standards (agent name/version, tool name, model provider, environment, tenant/app)
Partner with SRE/observability teams to ensure production-grade monitoring, alerting, and operational readiness
Collaboration with Teams (10%) Collaborate with transformation teams and business stakeholders to understand requirements and tailor AI agents to specific domains.
Work closely with AI platform teams to build scalable and cross-domain AI agents while ensuring end-to-end observability.
Integration & Deployment (10%) Build and maintain CI/CD pipelines for agent services and operations center components, including automated testing and deployment
Automate onboarding for new agent use cases (templates, scaffolding, configuration checks)
Drive best practices for secure, scalable, and cost-effective agent deployments
Continuous Learning (10%) Stay updated with the latest advancements in AI and machine learning technologies and integrate these into existing or new AI agents.
Conduct thorough testing and validation to ensure the reliability and accuracy of AI agents and solutions.
Qualifications
Key Skills/Experience Required Minimum Qualifications:
Education: Bachelors in Computer Science, AI/ML, Data Science, or a related field.
Experience: 35+ years of software engineering experience; 1+ years building and observe AI/ML or GenAI applications preferred
Required Expertise
Hands-on experience with agentic frameworks (Crew.ai, LangChain, Semantic Kernel, AutoGen, or similar)
Proficiency in Python (primary) and familiarity with APIs/microservices patterns
Strong experience with RAG patterns (embeddings, vector search, retrieval evaluation, chunking strategies)
Experience with cloud environments (Azure/AWS/GCP) and containerized deployments (Kubernetes/AKS/EKS)
Familiarity with observability fundamentals (logs/metrics/traces) and production troubleshooting
Experience building internal developer platforms or operational consoles (agent registry, run tracking, dashboards)
Familiarity with OpenTelemetry, distributed tracg, and telemetry pipelines
Experience with Azure AI Search / vector databases, prompt/version management, and evaluation frameworks
Knowledge of Responsible AI practices: data handling, safety guardrails, audit trails, and redaction strategies
FinOps exposure: token/GPU cost optimization and chargeback/showback reporting
Technical Proficiency: Agent orchestration design (planning, tool execution, memory, RAG), Strong engineering discipline: testing, versioning, CI/CD, automation, Operational mindset: reliability, debuggability, and incident response support
Problem-Solving: Ability to translate business challenges into technical solutions.
Collaboration Skills: Effective at working within cross-functional teams.
Agility: Flexibility to adapt to changing requirements and new technologies.
Communication Skills: Capable of explaining complex technical concepts to non-technical stakeholders. .
