注册一亩三分地论坛,查看更多干货!
您需要 登录 才可以下载或查看附件。没有帐号?注册账号
x
还记得当年解读analytics, 解读的DS的时候,今天来解读ML eng,感觉也是心中的兴趣在扩大和转移。是哪家厂子的job desc并不重要,其实大部分都类似
基本分三方面需求:
1. 要看得懂模型(有一定建模能力) 比如传统ML面 ,倒不见得需要是建模专家,因为一般是有research scientist来配合的
复习套路明确,但不容易
2. 要能码:比如leetcode那种coding面
复习套路明确,出苦力即可
3. 要能生产(at scale):包括系统设计和机器学习系统设计
系统设计:复习套路明确,
机器学习系统设计:不容易,没搞过估计非常难以准备。不过这条也只是plus,不是required,毕竟市面上真的做过生产规模系统的也就那么几家大厂
暂时还总结不好套路。可以读大厂的论文,但是这种系统级别的细节,只读论文估计也杯水车薪,一问就挂。
最近感觉ML infra是个有趣的方向,虽然并不懂很多。大厂小厂(or rather 好厂 vs 不怎么好的厂)之间差距已经可以多达5年。。。。
这条在老年以上级别会很重要
Infra / DE / production scale system Good understanding of distributed systems, data stores, data modeling, indexing and associated trade-offs - Experience with MapReduce, Hadoop, Spark, or TensorFlow is a plus.
Industry experience building and productionizing innovative end-to-end algorithmic or machine learning systems. Experience with productizing research/publications a plus.
Engineering fundamentals Full stack engineering experience, with strong system fundamentals Exceptional ability to work anywhere in the technical stack and good understanding of complex computer systems. Strong industry experience in architecture and development of scalable production quality backend systems
ML Machine learning, pattern recognition, large-scale data mining or artificial intelligence experience is a plus, but not required Good understanding of common families of models, feature engineering, feature selection and other practical machine learning issues, such as overfitting. Good understanding of data processing algorithms and/or common families of machine learned models. Effective at feature engineering and data analysis. Experience with Computer Vision, Image Detection/Recognition, NLP/NLU, Deep Learning a plus. Strong analytical thinking, good intuition with data, experienced with building products based on A/B testing 这条略奇葩了
|