How do you choose between DSML, DSA, SWE, and MLE?

i gotta say most of my life path is a result of randomness. my degrees are both in applied math, but w/ different application areas like finance and adversiting / recommendation algos. at the same time, i have tremendous interest in learning latest things, like recently the diffusion models and contrallable text generation. this insist in learning continuously made me decide to purse research related careers, thus, data or applied scientist in tech, or quant researcher in fiannce. for me, DS/AS at microsoft or amazon is the ideal option, it is close to production aka widespread impact, and the competition in the market pushes the team to use SoTA algo from the research world. i've also interviewed for other roles,
i. SWE, purly LC questions, i intewviewed w/ one FLAG company and got good results. however, i am not sure which team i will be working for, nor the tech stack... i am truly not a lover for frontend or backend dev for life. and there is only a very low prob that i can get matched w/ a ML prod+research team
ii. MLE, i interviewed w/ two MLE jobs and got good feedback. it was mostly about ML model optimization, i.e. using quantization, rewrite python code in C++, or other model compression methods. the main aim is to ship a model to production FASTER. i realize i am not attracted to that either
iii. DSA, i did not apply for any DSA positions, cuz i did not have that much time preparing for this postion specifc interview questions

How do you feel about your daily work

tbh, i am still happy and excited every morning. i go to office daily, and write my daily TODOs before getting my coffee. i work in advertising, which is defo the center of most tech companies. knowing the business is not less important than getting familiar w/ ML algos. i truly enjoying learning every details in the ads ecosystem. and my team consists of mostly applied scientists, each one is responsible for multiple models. there are weekly discussions about recent online experiments, comparing business KPIs of different algos; as well as discussions on latest innovating papers, sharing and comunicating w/ talented colleages def make my life much easier and happier! to sum up, the work life is basically reading papers, modifying codebase, training models, doing experiments, writing papers and a lot of intellectual discussions. since ads biz is mostly impactful during holiday seasons, it is also the busiest season, focusing on improving and experimenting existing models.

General advice for interview prep

i mostly googled each company, got their typical interview questions, and prepared accordingly. general patterns
i. ML basics, like in stanford cs224n, cs231n, cs229
ii. LC easy and medium questions
iii. SQL
iv. stats basics, esp testing related
v. core ML module impl, like attention, convolutional layers

Microsoft specific - almost useless referrals

sadly employee referral is nearly useless at microsoft, the prob of getting an interview from portal referral is basically the same as applying yourself directly
tho messaging the hiring manager is indeed effective, it is not happening a lot, since most people will only do so for people they have worked with

random things

learning is always fun, outside of work, i am recently exploring models related to autonomous driving and facial recognition. multimodality generative models also unlock many new oppotunities.