Final decision: CMU ECE PhD vs. UT Austin ECE PhD

背景：港中大CS本硕，做ML in EDA。不想做硬件。计划是找教职（首选回本校），当然也不排斥找工作（BTW，我们组毕业去华为的包裹现在都高的吓人了。。已经比美国cadence这种高点了）

CMU的导师是一位很随和，不push（甚至要你push他）的美国教授。主要是做DFT+hardware security。现在已经不怎么活跃了，（可能也是不push的原因吧）。最近在尝试找co-advise，不过不是老师主动去问，自己找也太难了点吧，不知道有没有前辈说说经验。

Austin的导师是华人，我比较熟，就是一个领域（EDA），也算业界大牛，属于学术壮年？虽然push但口碑不错，应该不算harsh？

CMU
Pro
1. 也许更高的声誉？所以在港校/新加坡找教职更好找？至少我系主任不带犹豫的推荐CMU。
2. 私立学校，对学生更友好？这是我看很多国外论坛帖子选PhD的一个推荐原因。
3. 据说CMU ECE PhD 可以免费拿一个 ML Ms学位，这样毕业找工的下限高很多吧？
4. 其实更想和美国教授合作，感受下美国的research style，在austin的话都是华人可能和在香港没区别。
5. 教授不push

Con
1. 完全不懂DFT和hardware security，看了下paper，说不上讨厌，但也没有喜欢。而且不知道这两个方向找教职难度如何。
2. 教授不active，毕业的学生很少找教职的，基本就是去硬件公司做DFT了，这两年有两位女生做ML in DFT， intern也都是在大厂（FLAG）往ML方向走，不过可能因为是女生，不具有代表性。
3. 以前组里没人intern（不知道是不让还是咋地），这几年，可以自己申请intern了。

Austin
Pro
1. 继续一个方向的research，以后找教职更容易（consistent research)
2. 教授学术壮年，和很多公司有合作，所以很容易找intern，教授也会推荐学生去3. 毕业学生出路基本都比较好，不论是找教职还是找工，（在地里看到的DFT的包裹好像都比较低）
4. 给了四年的fellowship，算下来比CMU钱还稍微多点。(3200刀一月，还没算intern）
5. 现在就可以直接开始合作，教授给了四年毕业的可能性。（当然得自己的work足够solid,influential)

Con.

1. 也许会有点harsh？因为现在教授对我的期望有点高，本来我都决定CMU了，还是把fellowship给了我，然后承诺找AI的prof co-advise，还给了四年毕业的可能，让我有点诚惶诚恐，到时候做不出好的work就有点。。
2. reputation比不上CMU，尤其是我想更偏CS一点的话。（毕竟是CMU哪）

谢谢各位！回复有米！

The reputation may not be as good as CMU. At least in HK, our department head recommended me to CMU without hesitation.


Hi guys, I just received the two offers and I was recently really hesitating. I have talked with many students and professors from CMU and Austin. But still, I did not make a clear decision.

I graduated from the CS department and I was doing ML in EDA previously. Personally, I did not want to touch on something related to hardware/EE. My future plan would be to find a faculty (at least current plan) at Hong Kong/Singapore, but I am always open to the industry.

I just list my summary here:

CMU

Pro.

Maybe higher reputation, especially in CS domain?

Private university, more friendly to graduates?

We can receive a "free" ML ms by enrolling required courses, which may contribute to our future career.

I actually kind of tend to work with US professor since I am too familiar with research in Asian group.

The professor is nice and do not push you harshly.

Con.

I am totally a rookie in DFT (Design for Testing) and hardware security. Also, the two topics may be too "hardware" to attract me and I don't know the difficulty to find a faculty in both directions.

The professor is not active now, and only a few students went to faculty.

Previously, the professor did not encourage intern, but recently, intern seems to be acceptable.

Austin

Pro.

I can continue my research direction, which seems to be more consistent.

The professor is pretty influential recently, and he seems to be still rising.

It is easy to find intern, because many companies, i.e., Facebook, Google, Cadence, will ask the professor to recommend students.

A fellowship is offered, therefore, the stipend is a little higher.
. From 1point 3acres bbs
Con.

Maybe a little harsh (just a little, many people said the professor was also nice)

The reputation may not be as good as CMU. At least in HK, our department head recommended me to CMU without hesitation.

Thanks guys!



补充内容 (2021-3-15 15:57):
欢迎回复，回复加米啊！

Hands down cmu

明显cmu

wanduzile 发表于 2021-3-13 15:48
Hands down cmu

感谢回复 是说easy mode 去CMU么

cszhazha 发表于 2021-3-13 15:49
明显cmu

感谢回复！能说下原因么 因为不push+CMU么

崔泽 发表于 2021-03-13 00:04:16
感谢回复 是说easy mode 去CMU么

导师人品好不push这是最重要的 况且还是cmu 然后研究方面找个兴趣相符的年轻老师coadvise肯定没问题

CMU +1

回复就给米啊🆙🆙

楼主请问EDA和计算电路/计算电磁的关系是什么啊

Fendapig 发表于 2021-03-13 07:40:46
楼主请问EDA和计算电路/计算电磁的关系是什么啊

我不太了解关系是什么
不过eda就是在设计芯片的时候很多步骤是需要自动实现的 比如placement routing（这好像只是eda整个流程中很小的一部分），我们就是研究算法解决这些问题。
总的来说，由于大部分input基本上都抽象成了cs 里的data representation，所以很多同学完全没有hardware 的背景。