3 ms·
>it is really hard to make a leap in improvement in technical skills. So true. I think it's exactly because of this that many professionals tend to reach a lev
by he11ow 7y ago
>it is really hard to make a leap in improvement in technical skills.
So true. I think it's exactly because of this that many professionals tend to reach a level and then expand their knowledge at the margin.From today's vantage point, it's incredibly difficult to tell which jobs will be around in 10, 20, 30 years' time. That makes me feel like learning new skills isn't a choice, it's a way to stay in the game for the long run.
Except that I don't think that the way you learn a bunch of skills in college still works when trying to learn deeper into one's career. It's not just the time, but the relevance.
I spent the last two years picking up skills in natural language processing, and the way I went about it was very different than reading intro books (personally, I found intro books least helpful).I've written about it on Medium, in case it's of interest:
https://towardsdatascience.com/learn-nlp-the-practical-way-b854ce1035c4 https://towardsdatascience.com/learn-nlp-the-practical-way-b...
- christiansakai 7y agoYes it is. It is also hard to keep up in this expanding field. Like, keeping up with kubernetes is a full time job, keeping up with machine learning techniques is also a full time job, keeping with frontend is also a full time job. Best practices change all the time. CQRS was all the rage now people want to do away with Kafka. Thanks for the link, I'm reading it.
- cookiesboxcar 7y agoCare to elaborate on your comment about CQRS and Kafka?