Skip to content
jonathanhodak.com
Return to Mission Control

Selected system / personal and professional work

Built and used

AI-assisted platform and infrastructure engineering

I use AI-assisted engineering across application, service, data, and infrastructure boundaries, carrying work from implementation into running and validated systems.

The problem

Engineering work often crosses frontend, backend, data, integration, and operations concerns. The work is incomplete when AI accelerates one code change but leaves the surrounding system, deployment, observability, or maintenance disconnected.

Scope

AI-assisted workflows span full-stack applications and services as well as Linux and Docker infrastructure. Agents help with controlled remote execution for work such as provisioning and configuring machines, installing runtimes and services, deployment, failure inspection, and environment validation. The dedicated development sandbox increasingly acts as a control environment spanning repositories, tools, services, and infrastructure.

My role

My contribution

I use AI-assisted workflows across architecture and implementation, carrying problems beyond the repository boundary into infrastructure and operations. Engineering judgment and validation stay with me.

System design

Architecture and technical decisions

  1. Treat the application and its service boundaries as the owners of domain behavior, authorization, and authoritative data.

  2. Use retrieval and vector search to provide bounded context to AI workflows while preserving source provenance and system ownership.

  3. Use MCP to expose capabilities across a boundary; keep domain rules and application logic in their service layers.

  4. Use bounded infrastructure workflows to carry architecture decisions into a running system, with explicit validation and recovery boundaries.

System boundaries

Constraints and qualifications

  • Domain rules, authorization, and authoritative application data remain in their service layers.
  • MCP exposes capabilities across a boundary; it does not become the home for application logic.
  • AI-assisted engineering in my environment does not stop at the code boundary: infrastructure work extends through controlled remote execution and environment validation.
  • Autonomy should increase inside engineered boundaries; do not assume controls that have not been established for a particular environment.

Design approach

Guiding principles

  • Carry a problem through application behavior, APIs, data, operations, and delivery.
  • Treat observability, backups, and deployment as part of the system, not follow-up work.
  • Use AI to reduce the time from understanding a problem to delivering a complete, validated system.

Conceptual architecture

System views

AI-assisted engineering continues beyond the code boundary.

I build and operate infrastructure as well as software. Bounded AI-assisted workflows help carry work into controlled remote execution: configuring machines and services, deploying applications, inspecting failures, and validating environments. My dedicated development sandbox increasingly acts as a control environment across repositories, tools, services, and infrastructure.

Infrastructure — conceptual system topologyAn architectural illustration across application, service, runtime, and operations, not live telemetry or an inventory of my private environment. Execution mode: once. Components: Application (React · full stack); API and service layer (Domain behavior · rules); MCP capabilities (Explicit integration boundary); Identity and access (OAuth · authorization); PostgreSQL and vectors (Structured · retrieved data); Agent runtime (Bounded execution); Workflow orchestration (Dependencies · failure paths); Containers and Linux (Runtime · isolation); Network and ingress (Routing · reverse proxy); Observability (Logs · metrics · traces); Backups and recovery (Restore · continuity); Deployment (Release · operate). Ordered stages: Application; API and service layer; MCP capabilities and PostgreSQL and vectors; Identity and access and Agent runtime; Workflow orchestration; Containers and Linux; Network and ingress; Observability; Backups and recovery; Deployment. Connections: Application to API and service layer; API and service layer to MCP capabilities; MCP capabilities to Identity and access; API and service layer to PostgreSQL and vectors; PostgreSQL and vectors to Agent runtime; Agent runtime to Workflow orchestration; Workflow orchestration to Containers and Linux; Containers and Linux to Network and ingress; Network and ingress to Observability; Observability to Backups and recovery; Backups and recovery to Deployment. Optional conditional paths are shown but do not run in the illustrated sequence.ApplicationReact · full stackAPI and service layerDomain behavior · rulesMCP capabilitiesExplicit integrationboundaryIdentity and accessOAuth · authorizationPostgreSQL andvectorsStructured · retrieveddataAgent runtimeBounded executionWorkfloworchestrationDependencies · failurepathsContainers and LinuxRuntime · isolationNetwork and ingressRouting · reverse proxyObservabilityLogs · metrics · tracesBackups and recoveryRestore · continuityDeploymentRelease · operate
  • idle
  • incoming
  • active
  • outgoing
  • settled

Conceptual sequence · not live telemetry.

Execution sequence

  1. 01
    ApplicationApplication → API and service layer
  2. 02
    API and service layerAPI and service layer → MCP capabilities · API and service layer → PostgreSQL and vectors
  3. 03
    MCP capabilities + PostgreSQL and vectorsMCP capabilities → Identity and access · PostgreSQL and vectors → Agent runtime
  4. 04
    Identity and access + Agent runtimeAgent runtime → Workflow orchestration
  5. 05
    Workflow orchestrationWorkflow orchestration → Containers and Linux
  6. 06
    Containers and LinuxContainers and Linux → Network and ingress
  7. 07
    Network and ingressNetwork and ingress → Observability
  8. 08
    ObservabilityObservability → Backups and recovery
  9. 09
    Backups and recoveryBackups and recovery → Deployment
  10. 10
    Deployment
Conceptual execution · explicit dependencies · not live telemetry

Intended effect

Intended effect

Experienced engineers can use AI to compress the distance between identifying a problem and delivering a complete production system, carrying the work across design, implementation, operations, and validation.

Technologies and components

  • TypeScript and React
  • React Router and Node.js
  • PostgreSQL and Supabase
  • RAG and vector search
  • MCP and OAuth
  • Docker and Linux
  • Vercel and deployment infrastructure
  • Monitoring and backups