SDV-scale software growth
Code size and change frequency continue to grow as OTA, connected services, ADAS, and software-defined vehicle programs expand.
Consulting
Automotive System/SW Development
AI Agent Consulting
We help apply AI agents such as Claude Code and Codex to SDV-scale automotive software development while respecting A‑SPICE, ISO 26262, SOTIF, functional safety, and human review constraints.
Overview
Using an AI agent is not the same as applying it safely to automotive development
The focus is not general AI tool adoption.We define where agents can assist code and artifact work, where practitioners must review, and what evidence must remain for standards-compliance activities.
Why Now
SDV development needs speed and safety at the same time
Automotive software is expanding into large-scale SDV systems through OTA, connected services, ADAS, and fast release pressure.AI agents can help, but only when the development and review structure accounts for standards, safety, and evidence.
Time-to-market pressure
Manual-only development, review, and documentation workflows are increasingly difficult to sustain under faster release cycles.
Limits of generic AI adoption
Using Claude Code or Codex in the same way as web or mobile development is not enough for automotive software engineering.
Framework
Design agent selection, sub-agents, Skills, and review rules together
We separate where LLMs are useful from where deterministic rules, templates, checklists, automation, and practitioner review must control the work.
Review Claude Code, Codex, Local LLM, or enterprise AI platform options against security, language, and standards constraints
Separate agent-assisted activities from activities that must remain under practitioner responsibility
Define AI output usage, review criteria, traceability, and evidence practices for A‑SPICE, ISO 26262, and SOTIF
Design sub-agents and Skills for requirements, design, code, testing, safety analysis, and review evidence workflows
Compare SaaS agent, internal platform, and Local LLM adoption models where security constraints require it
Scope
Support both source code and engineering artifacts
Automotive AI agent adoption should support not only code assistance but also requirements, design, verification, safety, review evidence, and document artifact workflows.
AI agent selection strategy
Decide where Claude Code, Codex, Local LLMs, or internal AI platforms fit based on project constraints.
Code development support
Define allowed use, restricted areas, human review criteria, tests, and evidence for AI-assisted source code.
Artifact authoring support
Support drafting and quality checks for requirements, design, verification, and safety artifacts while practitioners approve the final result.
Sub-agent and Skill design
Package repeatable tasks such as requirement quality checks, test case drafts, review checklists, and traceability cleanup into Skills.
Deliverables
Operating assets that remain in the project
The consulting output should remain as repeatable operating criteria, not only concept documents.
Agent/Skill adoption map
Map applicable agents and Skills to A‑SPICE, ISO 26262, and SOTIF activities, inputs, outputs, and review responsibilities.
AI coding agent rules
Define allowed scope, review criteria, prohibited use, and evidence practices for Claude Code, Codex, and Local LLM usage.
Artifact support templates
Provide prompts, checklists, and templates for AI-assisted requirements, design, verification, and safety artifacts.
Project adoption roadmap
Define pilot scope and transition plans so developers, quality teams, and safety teams operate agents and Skills under the same rules.
Apply AI agents to automotive development with standards, functional safety, and security constraints built into the operating structure.