2019 Kaggle Career Con Day 1 Recap

Today was the first day of the second annual data science career convention on Kaggle.com, a digital event focused on helping users land their first data science job. The following is a recap and summary of the sessions and my main takeaways and criticisms:


Session #1: Coding Workshop Part I: The Basics of Rest APIs

The first session of each day is devoted to a "coding workshop" aimed at teaching users how to get a data science API up and running. Starting off with a technical lecture that may or may not be applicable or even understandable to many viewers was probably a poor choice. The lecture was supposed to cover the basics of what REST APIs are and how to design one, but a combination of poor audio quality and lecture delivery made this session difficult to watch. This content could have been presented much more cleanly and concisely in a well-formatted video tutorial that was recorded instead of delivered live (something like a Kaggle Learn mini series.).

If you missed this session you didn't miss much.

Session #2: Career AMA with Dr. Andrew Moore

The second session of the day was a Q and A session with the head of Google Cloud AI. This session involved a lot of questions on a wide variety of topics so, so I'll just list the most important takeaways.

Learning by doing is of paramount importance. In terms of skills, knowing algorithms, probability and statistics is very important (Note, this is a very "Google" response.).

When hiring, he looks for people interested and curious enough that they've built things. You don't need to feel like you are credentialed or an expert to try things. It is important to find like-minded people, preferably people more skilled than you to learn from. You will learn much faster if you seek help.

You don't need a PhD or masters to be taken seriously in data science. Companies are most concerned about 1. good communication skills 2. the ability to work in a team and 3. a history of successfully completing/delivering projects.

This was a very productive session with some nice insight that attendees should find helpful and heartening.


Session #3: Panel of Working Data Scientists from Different Learning Backgrounds

The third session of the day was a Q and A session with a panel of early career data scientists on their backgrounds and experiences getting their first data science job. The first 15 minutes of the session were not particularly useful as they each just described their personal backgrounds. The upshot is that you can come into data science from a wide variety of fields and learning channels. The biggest takeaway from this session was that communication is extremely important in data science and you'll likely spend much more time talking with stakeholders and doing relatively mundane tasks like data cleaning than you will writing key code or doing machine learning. It was also highly recommended that you look out for mentors and join communities to learn faster. 

Session #4: Demystifying the Transition from PhD Student to Data Scientist

I skipped this session since I'm not a PhD student.


Session #5: Why Data Science Skill-Building Might Be Holding You Back

Session 5 was a Q and A session with Dan Becker, the head of Kaggle's learning content. Ironically, Dan's main point in the session was that data science learners with the goal of getting a job as quickly as possible usually end up ahead of more methodical long-term learners in terms of knowledge and career advancement. By jumping in and doing projects and getting a job as soon as possible, those looking for short-term advancement with minimal knowledge will ultimately start learning from others and on the job sooner, allowing them to advance further, faster than those who spend too much time studying. This is difficult but necessary advice that runs counter to the practical wisdom that slow and steady wins the race. Don't be slow and steady: dive into projects that interest you as soon as possible even if you don't know what you are doing. Learn what you need to know on the fly, then try to use those projects to get a job that interests you.

Other bits of practical advice from this session included: do something memorable in interviews to distinguish yourself ("wear a yellow scarf"), use visuals to draw attention in projects and put them front and center and don't busy yourself doing things you don't need to do to avoid doing things you are supposed to do or should do (don't procrastinate.). It is good to have a "T-shaped" skill set: a broad range of general knowledge and one area were you have deeper expertise.

In terms of applying to jobs, don't filter yourself out of positions by paying too much attention to job skill and experience requirements. People also overestimate the amount of complex math and often programming that data scientists do; those involved in ML research at top tech companies like Google might need high level math and algorithm skills, but that is not required for all (or even most) data science jobs.

This was a productive session with many useful tidbits of advice.

Session #6: Job Search Slack Mixer

This session was supposed to be an opportunity for job seekers to connect and chat on the Career con slack channel. Unfortunately there was no obvious place for users to congregate and talk. Kaggle made the session a sort of competition to make new chat channels to attract as many users as possible. I feel like this resulted in confusion and an unproductive splintering of the userbase into many different channels. This session would have gone better if Kaggle provided a handful of pre-defined interest areas for users to congregate in and talk, perhaps led by moderators/instigators from Kaggle, instead of making it a competition.


Day 1 of Kaggle Career Con had some productive discussions that were, unfortunately, book-ended by two relatively unproductive sessions at the beginning and end of the day.