AI Efficiency Architect (Intelligent R&D Productivity Expert)_BCSC

Location: Wuxi, Jiangsu, cn

Company Description

Bosch China Innovation and Software Development Campus

博世创新与软件开发中心

    博世创新软件开发（无锡）有限公司针对交通出行的电动化、自动化、互联化、个性化，提供面向未来的创新技术和前沿解决方案，加速针对中国市场的技术战略的实现和发展。博世软件中心主要发展方向包括智能网联汽车、智能座舱、自动驾驶、车路云协同、工业4.0、人工智能大数据、智能家居、嵌入式软件服务。博世中国创新与软件开发中心以软件为客户赋能，在汽车自动驾驶、氢燃料电池、重卡电驱动桥、多合一的控制器等多个领域取得创新研发成果。

Job Description

Role Overview
You will define and drive the AI-driven efficiency program for software development at BCSC. The core goal is to replace repetitive, rule-based daily tasks with AI Agents + Skills – specifically within Truck ADAS, Chassis & Powertrain embedded software development. You will start with the overall productivity architecture, identify automatable workflows, gradually break business logic into executable “Skills,” and orchestrate Agentic workflows, ultimately achieving a step-change in R&D efficiency.

Key Responsibilities
1. Productivity Architecture & Roadmap
Analyze the full software development lifecycle (requirements → modeling → coding → testing → integration → calibration) for ADAS and Chassis/Powertrain domains. Identify high-effort, low-cognitive, standardizable daily tasks.
Design a two-tier “AI Agent + Skill” architecture: define Skill granularity, I/O standards, and fallback mechanisms; design agent decision-making and orchestration logic.
Create a phased roadmap with measurable goals (e.g., 30% reduction in time for test case generation, 2x faster calibration data analysis) and visible gains within 3–6 months.
2. Skillization & Scenario Implementation
Convert typical engineering actions into AI-callable Skills, for example:
Automatically parse ADAS scenario requirements → generate test cases
Chassis controller change impact analysis → auto-update interface docs + unit test stubs
Anomaly pattern detection in calibration data → produce diagnostic report
Build and maintain a Skill repository to enable reuse and reduce onboarding costs for business teams.
3. Agentic Workflow Development
Orchestrate multi-agent collaboration: e.g., “Requirements Change Agent → Impact Analysis Agent → Auto Test Agent → Regression Verification Agent”
Enable Agents to call the existing toolchain (Jira, Git, Matlab/Simulink, Vector tools, Jenkins, etc.)
Design a human-in-the-loop interface: critical decisions require manual approval; routine tasks run fully automatically.
4. Business Alignment & Adoption
Work closely with UAES ADAS, Chassis & Powertrain development teams to understand real pain points under AUTOSAR, ISO 26262, ASPICE.
Collaborate with functional safety and quality teams to ensure traceability and compliance of AI-generated artifacts (code, test cases, reports).
Train engineers on using AI productivity tools, collect feedback, and iterate on Skills.

Qualifications

Must-Have Experience
5+ years in software development or R&D productivity, with at least 2 years focused on AI-powered engineering (e.g., code generation, auto test generation, defect prediction).
Familiar with ADAS or Chassis/Powertrain embedded software development processes. Experience with UAES or similar Tier-1 suppliers is a strong plus.
Proven track record of delivering AI Agent / Copilot / workflow automation projects (beyond just API calls), with measurable efficiency improvements.
Technical Skills
Solid understanding of LLMs (GPT-4, Claude, DeepSeek, etc.) and their application patterns: prompt engineering, RAG, function calling, multi-agent frameworks (LangChain, Semantic Kernel, or similar).
Basic to intermediate programming (Python/Java) – able to write Skill adapters and orchestration scripts.
Familiarity with at least two embedded development tools: Simulink, CANoe, Davinci, Trace32, etc.
Plus: Knowledge of ASPICE, ISO 26262, AUTOSAR architecture.
Core Mindset
Engineering mindset: ability to turn vague “efficiency ideas” into measurable, implementable Skills.
Business-driven: focus on solving real, tedious pain points for ADAS/chassis engineers – no tech for tech’s sake.
Persuasive & coaching ability: willing to change team habits, backed by data.