kind: AgentProfile
apiVersion: profile.github.io/v1
metadata:
  name: Junming Wu
  labels:
    domain:
      - ai-infrastructure
      - inference-serving
      - agent-systems
      - kubernetes
spec:
  role: AI infrastructure and agent systems builder
  optimizedFor:
    - shipping AI systems that are deployable, observable, and maintainable
    - turning rough automation ideas into reliable workflows
    - keeping inference and serving paths boring in production

runtime:
  summary: I work around the parts of AI systems that decide whether a demo becomes a service
  areas:
    - inference and model serving infrastructure
    - AI applications, agents, and tool-using workflows
    - Kubernetes-based deployment and operations
    - observability with metrics, logs, alerts, and incident context
    - automation that removes repeatable operational work

capabilities:
  build:
    - AI app prototypes with real operational paths
    - agent workflows with clear tools, context, and constraints
    - internal automation for infrastructure and reporting
  operate:
    - Kubernetes workloads
    - inference-serving environments
    - monitoring and alerting flows
  improve:
    - deployment reliability
    - observability signal quality
    - debugging speed during incidents

toolbox:
  platforms:
    - Kubernetes
    - Linux
    - GitHub
  observability:
    - Prometheus
    - VictoriaMetrics
    - Alertmanager
    - logs and dashboards
  ai:
    - LLM applications
    - agent design
    - tool orchestration
    - inference and serving workflows

operatingMode:
  default: pragmatic
  preferences:
    - make systems understandable before making them clever
    - prefer boring production paths over fragile magic
    - automate the second painful repetition
    - keep alerts actionable and dashboards tied to decisions
    - treat agent behavior as an interface, not a vibe

currentFocus:
  - AI infrastructure for inference and serving
  - agent workflows that can survive real-world context
  - Kubernetes-native operations for AI workloads
  - observability patterns for faster troubleshooting

collaboration:
  worksBestWith:
    - clear goals
    - measurable failure modes
    - enough context to understand production constraints
  returns:
    - practical implementation plans
    - operationally aware systems
    - automation with fewer hidden assumptions

constraints:
  - no mystery infrastructure
  - no dashboard without a decision behind it
  - no agent workflow without explicit tools and boundaries
  - no production change that cannot be observed

status: available for building practical AI systems