Technical Program Manager (Silicon Programs)
virtual
· 难度 3/5
流程通常从 recruiter screen 开始,强调职责范围、跨职能匹配度,以及复杂排期经验。之后候选人反馈会和一位 hiring manager 以及来自 engineering 和 operations 的多位 stakeholder 面试,重点在规划严谨度、风险管理和沟通。面试 loop 经常包含情境题,围绕里程碑执行、依赖跟踪,以及在硬件和软件团队之间的升级策略。周期通常是几周,决策与跨职能面试官之间的对齐程度相关。
由 AI 翻译,查看原文
- Describe how you build and maintain an end-to-end silicon program plan across architecture, design, verification, and validation milestones.(描述你如何建立并维护一份端到端的 silicon 项目计划,覆盖 architecture、design、verification 和 validation 里程碑。)
- Tell me about a time you resolved conflict between two teams with competing priorities. What was the outcome?(讲讲一次你解决两个优先级冲突团队之间矛盾的经历。结果如何?)
- If a critical deliverable slips and threatens a tape-in or launch milestone, what steps do you take in the first 48 hours?(如果一个关键交付物延期,威胁到 tape-in 或 launch 里程碑,前 48 小时你会采取哪些步骤?)
- How do you ensure transparency and accountability without creating a blame culture on high-pressure programs?(在高压项目中,你如何确保透明和可追责,同时又不制造甩锅文化?)
Graphics Driver Software Engineer
virtual
· 难度 4/5
流程通常从 recruiter 电话开始,然后是一轮技术面试,考察底层编程、OS 概念和调试。候选人经常会经历一轮更长的线上 loop,多位工程师参与,讨论内容覆盖 C 或 C++ 深度、并发、性能 profiling,以及 GPU pipeline 基础。有些团队会加入一个实操环节,例如分析一个 bug 场景或对代码进行推理,即使不要求正式的 take-home。整体周期常为几周,跨时区协调也可能拉长排期。
由 AI 翻译,查看原文
- Walk through how you would diagnose a GPU hang or TDR-like event and what signals or logs you would collect first.(讲讲你会如何诊断 GPU hang 或类似 TDR 的事件,以及你会优先收集哪些信号或日志。)
- Explain common causes of race conditions in driver code and how you would design synchronization to minimize contention.(解释 driver 代码里 race condition 的常见原因,以及你会如何设计同步来尽量减少 contention。)
- Tell me about a time you improved performance or reduced latency in a system-level component. What did you measure and how?(讲讲一次你在系统级组件里提升性能或降低延迟的经历。你测了什么指标,怎么测的?)
- How do you balance correctness, performance, and maintainability when a change is needed late in a release cycle?(当 release 周期后期必须做改动时,你如何平衡正确性、性能和可维护性?)
Silicon Design Verification Engineer
virtual
· 难度 4/5
流程通常从 recruiter screen 开始,重点在岗位匹配、工签/授权以及时间安排预期。候选人反馈会有一到两轮技术电话或视频面试,深入验证基础能力和过往项目深度。之后通常是线上 panel loop,多位工程师加上一位 hiring manager 参与,混合 SystemVerilog/UVM 讨论、调试场景题和 behavioral 问题。周期通常持续几周,反馈节奏会受团队时间和 HC 审批影响。
由 AI 翻译,查看原文
- Explain how you would structure a UVM testbench for a new block and what components you would prioritize first.(解释你会如何为一个新 block 搭一个 UVM testbench,以及你会优先做哪些组件。)
- How do you debug an intermittent failure in a constrained-random test and determine whether it is a testbench issue or a design bug?(你如何调试 constrained-random 测试中的间歇性失败,并判断是 testbench 问题还是 design bug?)
- Describe a time you found a critical bug late in the cycle. How did you communicate it and drive resolution?(描述一次你在周期后期发现关键 bug 的经历。你如何沟通并推动解决?)
- If coverage is behind schedule, how would you decide what to target next and how would you negotiate tradeoffs with design?(如果 coverage 落后于计划,你会如何决定下一步优先补什么,以及如何和 design 协商取舍?)