The second day of the second annual data science career convention on Kaggle.com concluded today. The following is a recap and summary of the sessions and my main takeaways and criticisms:
Session #1: Coding Workshop Part II: How to Make an API for an Existing Python Machine Learning Project
Day two kicked off with a continuation of the coding workshop started on day one, focused on teaching users how to launch a data science model as an API. Like day one, I generally found the coding session to be difficult to listen to and follow due to a combination of poor audio and fast-moving, sometimes unorganized and cloying lecture delivery. This session did have some more interesting and useful things in the live coding section, such as a discussion on how to save trained Python models and objects, but I feel that these sessions could be presented better as prerecorded videos and notebooks.
Overall an improvement over day one, but still too difficult to follow.
Session #2: Tips for Getting Your First Data Science Interview with No Personal Connections
This 30 minute session consisted of a 20 minute presentation with some job search tips followed by 10 minutes of questions and answers.
Session #4: How Data Scientists Can Build an Online Brand, Even If Self-Promotion Isn’t Your Thing
This session consists of a short bio on Siraj Raval, a YouTube ML content creator, followed by some general questions on content creation and establishing an online presence. For Siraj, when it comes to content creation, three things are key: 1 consistency, 2. being data driven and 3. having a mission/purpose beyond promoting yourself. It is also important to have fun because content creation is a lot of work and you won't be able to be consistent if you don't enjoy it.
This was an good session if you are interested in content creation, but otherwise it wasn't as useful and perhaps didn't fulfill its stated goal. It didn't really address the point of building on online brand "Even If Self-Promotion Isn’t Your Thing" since content creation is, in a sense, form of self-promotion, as is general social media use. It would have been nice if they focused a bit more on people who aren't content creators and don't have a social media presence.
Session #5: Career AMA with Jeff Dean
Session #2: Tips for Getting Your First Data Science Interview with No Personal Connections
This 30 minute session consisted of a 20 minute presentation with some job search tips followed by 10 minutes of questions and answers.
The key takeaway from the lecture was that data science is a broad field so the responsibilities of a given data scientist job posting can vary greatly. It is important to do research on different roles and dig in to target those that fit your skills and interest, tailoring your application materials to the position, rather than just applying to every data science role you find.
In terms of interviews, it is important to know your resume since it will likely come up and prepare yourself for challenging questions by practicing. Try to enjoy the process as best you are able.
Unfortunately the presenter's mic cut out during the Q and A session so not much of value came out of it, which is a shame because there were some good questions.
Session #3: Live Portfolio and Resume Analysis with Data Science Hiring Managers
The live resume analysis session involved 3 hiring managers looking at 3 different resumes submitted by Career Con attendees. All three of the resumes chosen were candidates with higher education in tech fields and 2 out of 3 either had a PhD or were earning a PhD. The fact that resume details are so specific to one person combined with the fact that they focused heavily on people from highly academic backgrounds made this session feel less valuable than it could have been and perhaps disheartening for viewers without traditional higher education. In addition, some of the advice conflicted, which shows that companies and hiring managers look for different things and that there is often no right answer as to how to structure a resume.
Session #3: Live Portfolio and Resume Analysis with Data Science Hiring Managers
The live resume analysis session involved 3 hiring managers looking at 3 different resumes submitted by Career Con attendees. All three of the resumes chosen were candidates with higher education in tech fields and 2 out of 3 either had a PhD or were earning a PhD. The fact that resume details are so specific to one person combined with the fact that they focused heavily on people from highly academic backgrounds made this session feel less valuable than it could have been and perhaps disheartening for viewers without traditional higher education. In addition, some of the advice conflicted, which shows that companies and hiring managers look for different things and that there is often no right answer as to how to structure a resume.
The main point of consensus was that projects are very important.
Session #4: How Data Scientists Can Build an Online Brand, Even If Self-Promotion Isn’t Your Thing
This session consists of a short bio on Siraj Raval, a YouTube ML content creator, followed by some general questions on content creation and establishing an online presence. For Siraj, when it comes to content creation, three things are key: 1 consistency, 2. being data driven and 3. having a mission/purpose beyond promoting yourself. It is also important to have fun because content creation is a lot of work and you won't be able to be consistent if you don't enjoy it.
This was an good session if you are interested in content creation, but otherwise it wasn't as useful and perhaps didn't fulfill its stated goal. It didn't really address the point of building on online brand "Even If Self-Promotion Isn’t Your Thing" since content creation is, in a sense, form of self-promotion, as is general social media use. It would have been nice if they focused a bit more on people who aren't content creators and don't have a social media presence.
Session #5: Career AMA with Jeff Dean
The final session of the day was an interview with Jeff Dean, SVP of Google AI. This session was focused more on Jeff's background and interesting data science ideas than than the data science job search process, touching on ideas like AutoML and his desire to see models that are able to optimize over many problems at once. Ideally, we'd train one large model that can answer many or even all the questions we want answered instead of training models on extremely narrow, specific tasks like we do now.
The main takeaways from the session in terms of a job search were: a strong fundamental understanding of computer science basics is important and it is good to surround yourself with people you can learn from who know things that you don't.
When asked whether people from non-traditional educational backgrounds can work in ML research, he suggested it is possible and that most people in their residency program succeed, but his example of a "less traditional" person in the residency program was someone with a physics masters coming from industry instead of PhD. I think the message here is that those with nontraditional backgrounds need to temper their expectations in regards to doing cutting edge ML at tech giants. To echo advice from yesterday's session: you'd be better served by starting somewhere now than trying to learn enough on your own to jump right to the top.
The main takeaways from the session in terms of a job search were: a strong fundamental understanding of computer science basics is important and it is good to surround yourself with people you can learn from who know things that you don't.
When asked whether people from non-traditional educational backgrounds can work in ML research, he suggested it is possible and that most people in their residency program succeed, but his example of a "less traditional" person in the residency program was someone with a physics masters coming from industry instead of PhD. I think the message here is that those with nontraditional backgrounds need to temper their expectations in regards to doing cutting edge ML at tech giants. To echo advice from yesterday's session: you'd be better served by starting somewhere now than trying to learn enough on your own to jump right to the top.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.