If you have any doubts about the business case for teamwork, just take a look at the Cinderella story of the first competitive Kaggle Team in the Philippines.
For those not familiar with Kaggle, it’s a site popularly known as the Olympics of data science – an arena where the top data scientists from academia and industry compete to write the best predictive algorithms. The stakes are high – the competition data comes from real companies—and the prizes run the full range from bragging rights up to $500,000.
Our story starts in May, 2015. I was approached by some of the largest tech companies in the Philippines to train their next generation of Data Science talent. The project brief? Grow fresh graduates from the local university system into bonafide data scientists in just 5 short weeks.
I was initially hesistant – 5 weeks of teaching was barely enough to cover foundational data science skills let alone math, stats, coding, big data and machine learning. I realised that there was only one way that the project could be realistic – I needed the students to be amongst the best graduate talent in the country. I sent the HR departments of each client off to recruit the top brains in the nation and immediately started work on the course syllabus.
And a course syllabus it was. A gruelling 27 day, 8 hours-a-day bootcamp designed to sprint from ‘hello world’ to Hadoop in record time. I forwarded the completed course outline to a former teaching colleague at the Australian National University, who promptly informed me that:
- I was borderline crazy.
- The program was only a few contact hours short of a full 1 year Master’s degree program.
- “They’d better be damned smart”.
With my colleague’s words of caution echoing in my ears, I steeled myself for the beginning of the Bootcamp. On Day 1, I met my new students – highly motivated math and statistics graduates from the top universities in the Philippines. So far, so good.
Shortly after teaching initial concepts, I realized that these students were smart. Really smart. Kaggle smart. It was game time. Around Week 4 of the course, we officially formed our team (“PointSeerPH”) and enrolled in the Caterpillar Tube Pricing Kaggle Competition, a price prediction problem issued by the world’s largest manufacturer of construction equipment. Could we create a pricing algorithm that would place well enough to make us the Cool Runnings of Data Science?
As expected, our first submission was a let down. We entered the leaderboard in 1,069th place out of 1,333 entrants (we are “PointSeerPH”):
But then, we started to move up the leaderboard as the competition unfolded. Every day our ranking went up. We quickly learned how to work better as a team, dividing the workload, pooling computing power and having small groups experiment with different machine learning models.
So let’s cut to the chase – how did we do? The team of fresh grads, who didn’t even know how to code when they started are currently ranked 73 out of 1,333 teams. That puts them in the top 5% of the Kaggle leaderboard (and the competition isn’t even done yet!):
So how did a bunch of newly trained graduate underdogs sneak into the top 5% on Kaggle? Let’s look at the team dynamics at play that allowed a group of dedicated students to accomplish in a month what some data scientists with years of experience sometimes can’t:
- Low expectation: This might sound completely out of left field (especially when we live in a world of corporate metrics, strategic goals and deadlines), but when given an open ended problem, the absence of a defined target allowed us to explore and try out new theories and ideas. No one knew that we were competing in the competition except ourselves. This low expectation mindset was so liberating that everyone was able to ask questions and get answers without worrying about looking stupid in front of superiors.
- Competition: The idea of competing against the best data scientists from Silicon Valley energized the team, to the point that working on the problem felt more like a sporting event rather than work. There was also healthy intra-group competition of the best kind – everyone was helpful and communicative with each other, but there was certainly an air of wanting to be the person responsible for having our team move up another 100 places on the leaderboard. This really drove creativity and motivated everyone to create winning code.
- Having fun: Happy people are productive people! Spending 5 weeks in the same room brought everybody close together. So instead of banging heads against tables when something didn’t work out right, the release valve was a joke and a timeout with friends. This fun and supportive setting meant that team members powered through roadblocks and setbacks in situations where a solitary tech worker may have gotten frustrated and called it quits.
I hope that this team of young graduates can be a reminder to us all of the legendary power of teamwork!