The tricky part is setting up your pipeline: from .csv files to submission. This can take a day, or 30 minutes, depending on the contest. If feature engineering is required then this adds a lot of time to this process.
Time to create the first submission dropped drastically after a few competitions. I now have a small library of munging and ensembling scripts that I can quickly adapt to suit the needs. On the other hand, time spend on optimizing and staying inside top 10, increased too. For the KDD-cup I'ddo weekend long sprints for a few measly improvements. All in all I'd say I spend at least 8 hours per competition.
My background was front-end developer growing into analytics and dataviz more and more. I think it was on HN that I saw a link to learnpythonthehardway.org and I started from there. After reading "programming collective intelligence" I got more serious.
And your post is certainly an inspiration to me :) Thanks for documenting your journey. I'll make sure, one year down the line, I document something like this and hopefully will get some success on kaggle.
Awesome, thanks for sharing (your experiences and the book pointer). Definitely motivating to see your progression and I'll have to make some time to enter some competitions.
No, not yet. I'd love for Kaggle to be a place where analysts can earn a steady income, but I think that is something for the future. Right now even the top performers do not consistently win prize money, so they'll need a job on the side. If they are top competitors then this is usually a wellpaying job and so they are competing for fame and bragging rights, not prize money.
At my startup I do predictive analytics on whatever data I can get my hands on. There is an industry demand for fast and scalable solutions, which I fancy working on. Also Kaggle indirectly gives a lot of commercial opportunities and exposure. Perhaps in a year I'll reflect back on a modest start with ML as a career.
Also what was your background before starting to compete?