I work as a data scientist and have for a few years now for multiple companies -- I've worked on machine learning projects with business impact, and I've worked on lots of data analysis. Most of the time though, there's more need for data analysis. There's always a fair bit of frustration that comes with the role that never really gets resolved -- dealing with slow data infrastructure, too much data cleaning, too many data requests from too many stakeholders, being otherwise out of the loop once people stops needing data, etc. Over time I've felt the role to be more draining and frustrating than interesting. I also feel that a high level DS is less impactful than a high level Eng or PM. Also I feel that moving to other companies would be more of the same, but maybe if I put in the time to study and interview or even just switch teams I can find a role that is more true data science - which in my mind is more ML modeling and quantitative research. Meanwhile I have a side job -- it is not data science at all but it is in a interesting and new area. The job can grow to be more than a side job but then would require me to move to a different country. (For privacy this is all I can reveal about the nature of the side job). The pros are that it would be an adventure with potentially lots of upside, the cons are that I don't have as much experience in this area so would kind of be starting over and the pay may not be as high, and that I really prefer living in the US to the other country. There are times when I like being a data scientist a lot, I love working with numbers, I'm actually positive on Uber's future, and I like the bay area. Lately however the draining and frustrating parts of being a data scientist has been more acute for me, and I feel like I should do something about that since I still spend most of my day doing that. So what should I do? A - just keep doing my full time job as a data scientist and my side job until it becomes clearer which way to go B - study and interview to find a better full time job, all the while keeping my side job C - take a risk on the side job, move to the other country
Leave asap
thanks, can you give some rationale? There are some good parts to the job that maybe I'm not accurately reflecting right now.
Skill wise: 1. Show your skills for a role specific to deep learning. 2. Or Take interest in distributed machine learning with a focus to grow more on distributed systems side. If you are saturated with existing work domains, try increasing breadth on other things. Level wise: 1. If you are not a senior, move to FNG for roles inclined more towards research than application, where you get to develop models more than data cleaning. This means bump in your TC as well. 2. If you are a senior, show your technical prowess on modelling part. Create prototype and propose whole new project to the management, which you get to drive completely. This might mean sacrifice of time used for your 'side job'. About your side job: 1. Staying US and love for ML mean letting go your side job. Or keep your 'side job' a side job, let not it become the main job. 2. If you are financially fully backed up and want to try other jobs, go for your 'side job' and make it your main job. This looks difficult if you are married and have kids. But if you're passionate about it, then go for it.
Thank you for your comments. Regarding your level wise comments, why only move to FNG if I’m not senior? Also I think data scientists do about the same thing at FNG. I am senior though fwiw, the problem is I am overutilized as it is that I don’t have time to be proposing new projects. About the side job - part of the question I’m struggling with is if I should put in the extra effort for ML or should I just try something new altogether. I don’t have kids.
If you are a senior, it makes sense to stick to the same company to make it easier to climb ladders. You can move anywhere, not just FNG, but definitely an advantage if it is. Else you can try flourishing startups, but you've already experienced it. You might get good TC bump if you move to Lyft, as Uber folks will be directly useful for them. Other key options I see are Quora, LI and Airbnb. Extra efforts in ML are worth it if you are bound to read latest ML or DL papers every week as a part of your job. Otherwise, to continue, you have pretty much experience with data science in industry. Changing the domain altogether, like moving to hard core Distributed Systems, would be a big change.
I totally feeling you as a DS. Write a best case DS workflow and recommendations as it relates to your enviroment and see if you can fix it - you probs can’t. That will make you face the problem, find a solution, and share it. Once that doesn’t work, yeah, fuck DS, do something else. It’s broken and it’s getting shittier every year.
I like your point about facing the problem.
Well I’m not really thinking of becoming an engineer, but maybe an “applied researcher” or “ml engineer” or whatever the title is for what data science was advertised to be.
This
Had same thoughts, switched to ML engineering, happy now. But I knew from the beginning what I wanted to do, just did not know what a particular role implied.
The decision will be yours and yours alone. So ask only the questions that will help you make that decision. For now you are asking us to make a choice for you. That's not going to help you much because we don't have enough data about your predicament to make the right choice.
The stakeholder data request noise can be overwhelming and it is BI work. It can be painful if data scientists are also doing BI work. Try a place where there is only analysts and ML engineers. Analysts take on the BI workload and ML engineers take on the DS work.
That sounds good in theory - in fact Uber has both product analysts and data scientists. However, when we are understaffed then I have to cover for both. Also, I don’t think that product and eng necessarily see the distinction between analysts and scientists.
Try the applied scientist role in Amazon or MS. It might be what you're looking for.
Quit asap
In a similar spot to you, same feeling of being over utilized and not being able to drive strategic vision. From my perspective, the best strategies have been to say no to tasks, creating repeatable scripts or notebooks to automate mundane asks, and trying to teach others what the appropriate questions to ask are. I think Quora is better culturally about how we use data than others, but there are definitely data problems which won’t go away (poor data infra is def a big one). The one good thing is that people, esp PMs understand how to use data so we don’t get that many stupid requests. It’s also a smaller org so fewer levels to go through to propose projects Happy to dm if you wanna chat more
TC || 👉
300k