How to go from Data Science to ML/AI Engineering

SEI Investments Quanty
Oct 29 20 Comments

I currently have 2.5 YoE as a data scientist for a large company. Job involves running ML algorithms on data sets after cleaning/sanitizing them and working with business stakeholders to find requirements/etc. In a profit center of the company right now working on building new implementations.

I'm getting annoyed with the "70% of date science is clean-up" part and I want to do more SWE and coding. For example, the idea of building self driving cars or working on AI to launch shit into space sounds awesome.

Starting master's in computer science with ML spec next spring with Georgia tech but don't really know how to get into the space now (as opposed to latter). Also taking deeplearning.ai course on Coursera to get up to speed on Tensor flow since I just use SKLEARN now. Also only know python but gonna learn other languages soon.

Any advise appreciated! Thanks

comments

Want to comment? LOG IN or SIGN UP
TOP 20 Comments
  • Cisco / Eng 👨 🐻 🐖
    I did exactly what you’re trying to do. Best thing you can do right now is automate the data cleanup and build a really good ETL engine from scratch. That’s how I did it
    Oct 29 4
    • SEI Investments Quanty
      OP
      Data Clean-up isn't always so straight forward. For example, reading over documents to pull up important phrases of text for setting up ML models. We have lots of random projects which require unique solutions. Also we already have a full ETL time that handles all data loads and I admit I don't really know much about ETL other than ours works for putting large amounts of data into our DWH.
      Oct 29
    • Cisco
      🦙 Alpaca

      Cisco

      BIO
      I am the Alpaca. DM me if you want to cuddle.
      🦙 Alpacamore
      Lots of random projects, like 10? Automate one by one.
      Oct 29
    • SEI Investments Quanty
      OP
      It's like "Do X project in Y weeks then move onto the next". I have automated things that are easily automated, but don't know how I'd automate new projects I don't know about yet.

      Although one thing to point out is that this is more about me changing careers. I want to get a new job at a new company but don't really know how to get the pre requisite experience on my resume.
      Oct 29
    • HCA Healthcare levoflox
      Dont forget about data instrumentation and data monitoring. Learn about concept/model drift and how to detect it. Try and automate that process
      Oct 29
  • American Express D.BCooper
    Majority of companies hiring "data science" just end up having you scrub data and no idea what supervised vs unsupervised learning is.

    Python, etc are the new buzzwords. Interview with a few companies and they're not even far along enough with their data pipelines or data collection mediums (e.g., loyalty card) to start serious analysis.

    The other issue is there is a serious shortage of talent so B player companies can't hire anyone.
    Oct 29 0
  • Snapchat hippbutrr
    how about starting coding more? making more production code, huh?

    build your ML models and then deploy them end-to-end, ok?
    Oct 29 1
    • SEI Investments Quanty
      OP
      That's what I do. Doesn't mean I can get a job at Uber making self driving cars.
      Oct 29
  • LinkedIn zkmH23
    Software engineering skills are more important now if you already do ML as a data scientist. masters in ML will be useless. The only thing that may help is strong expertise in a particular domain, like vision/robotics/nlp if you want to get job in that domain. Not that many MS programs with domain specialization, but they exist.
    Oct 29 8
    • LinkedIn zkmH23
      To be honest, most candidates I interviewed who did masters in ML know shit about ML. Too much breadth without real depth. If you take courses on all the subjects that you mentioned - you won’t really understand any of these disciplines. Based on what you described you are doing now, you should be able to impress your interviewers with ML skills more than a typical Masters in ML graduate. Now, to be a ML engineer you need to also impress your interviewers with engineering skills. How to address scalability in training, are you scoring offline or online, how do you index your data, what is the latency of request? With DS background instead of doubling down on ML knowledge it is much better to invest in improving engineering skills.
      Oct 29
    • SEI Investments Quanty
      OP
      That's an interesting perspective I will definitely keep in mind. How would you recommend I navigate my career from here? Also, in case it's relevant, I'm continuing to work Full time and getting master's part time.
      Oct 29
    • LinkedIn zkmH23
      I’m not saying it is useless, but you don’t have infinite time to spend. It is better to spend time on engineering and or Masters in ML focus area, where they don’t try to teach you all of ML, but only a sub field of ML.
      Oct 29
    • LinkedIn zkmH23
      If you do ML masters make sure you have good emphasis on fundamentals: linear algebra, statistics, optimization, distributed ML, classical ML algos, neural networks. Don’t try to cover maximum number of modern application areas at once.
      Oct 29
    • LinkedIn zkmH23
      And don’t underestimate the importance of software engineering.
      Oct 29
  • HCA Healthcare levoflox
    I was a DS at a non-tech company that also did critical backend work and did a lot of production ops. I applied for several ML eng jobs. Got auto rejected without any interview from most except one. Ended up getting that job. Luck is huge. Shoot your shot.
    Oct 29 2
    • SEI Investments Quanty
      OP
      What got you auto rejected and how can I bolster my resume so that doesn't happen to me?
      Oct 29
    • HCA Healthcare levoflox
      3+ YOE at a non tech company without a cs degree or masters or phd is probably why I got auto rejected. I managed to get into a FAANG now so I doubt that will be an issue again. Sometimes you just need that big break
      Oct 29

Salary
Comparison

    Real time salary information from verified employees