Meet the Winners of Xeek’s Aera Energy Challenge

Another Xeek challenge for the books! This competition was in partnership with Aera Energy to study steam optimization for their operations. There were two prize pools for this challenge, one traditional Xeek pool and another to encourage participants from Colorado School of Mines. Learn all about the winners below.

First Place: Mr. Professor (Team), $12,000

The Mr. Professor team took first place with a fast and efficient CatBoost model. But what really stood out was their thinking around exploratory data analysis, taking the time to write good documentation of their code, and submitting ideas on how to deploy their code. Their model was great, but these extra steps make implementing and growing their solution that much easier.

The Mr. Professor team enjoys competing in all kinds of data science challenges, but they were particularly drawn to Xeek’s unique approach.

“We came across this competition, and the very first thing [we noticed] which we liked was that it was not only about the leaderboard,” said team member Anil Betta. “It included creativity and the entire data science lifecycle […] and our expertise helping [geoscientists] in real scenarios is very exciting for me. Our work is being utilized by someone.”

Taking part in competitions like Xeek’s “Steam Optimization and Other Oddities” goes beyond winnings and accolades. Betta said, “data science is an ever-evolving field. If we just do our office work, we won't be upskilling ourselves. And the expertise I get from competition is directly visible in my office — most [people have to] Google things that I know by heart because of competitions. It really helps me a lot in my day to day.”

Their advice to potential Xeekers is to really dig in and not get discouraged by the challenges that arise. It is a challenge, after all! “It takes time,” said Betta. “So just don’t give up.”

Second Place: Pavle Mihajlovic, $6,000

Pavle Mihajlovic came in second using a novel approach for matching models. The judges were unfamiliar with approaching production data with this method and were interested to see where this work could be taken further. His work also stood out as having a well-documented dive in the data. Having an outsider explain your data is a way of seeing the problem in a different light, which can lead to further insights.

Mihajlovic was interested in the Aera Energy Challenge because it was completely different from competitions he had done in the past. “I was looking for some machine learning competitions to participate in and what I liked about the Xeek challenge was that the data could fit into memory,” said Mihajlovic. “When you're running machine learning competitions and you're dealing with incredibly large data sets, ones that don't fit into memory, it becomes a bit of a hassle sometimes to work with. So, I enjoyed the fact that it was sort of a smaller data set. You could understand what's going on across the entirety of it. I remember [finding the competition] and thinking oh, this is interesting. I definitely could do some interesting work here.”

“What also attracted me to it was the grading schema,” he continued. “With most data science competitions, you just want to optimize a single metric. And anybody that does data science for a living knows that it's never as simple as just minimizing or maximizing an arbitrary metric — you need to be able to think about who is going to consume it, how they're going to consume it. How can it become usable? Is it interpretable? A lot of other competitions are missing out on that aspect. [The Xeek challenge] reminds me more of a fully-fledged data science project. The manner in which it’s graded allows you to put a lot more thought into the solutions.”

Third Place: Thomas Martin, $3,500 / First Place Colorado School of Mines Prize Pool: $2,500

Thomas Martin from Colorado School of Mines won third place in the general prize pool and first place in the CSM prize pool. His solution stood out because he used geologic logic to put limits on what the model could do. These geologic insights make for more realistic outputs and provide confidence that the model can be transferred to other fields of operation for the industry.

Martin’s advice to anyone thinking about taking part in a Xeek challenge? “I would say just do it,” said Martin. “It's hard to go wrong. Even if you only spend five hours on it and you might not be in the top 10, I can probably guarantee you're going to learn something new. I learned something new. I know there are a couple takeaways from this project I'm excited about besides the monetary stuff. So, if you're interested in data science, if you're interested in real data, and you're interested in how all of those things work, I would say why not? It's easy to get started and it is a really great place to learn.”

Second Place Colorado School Mines Prize Pool: Owen Huff and Rosie Zhu, $1,000

The CSM team of Rosie Zhu and Owen Huff was picked because they showed how they came up with their solution in the submission. The team experimented with several different models and discussed why they did or did not work. Sometimes knowing where not to dig is just as important as the treasure map.

As students, Huff and Zhu have busy schedules and took advantage of Xeek’s resources to help them navigate the challenge. “All the questions we had during the office hour really helped,” said Zhu, “and people asking questions in the forum was helpful as well. The people at Xeek were really responsive to all the questions I had — I had a lot of questions! [The challenge] was hard, but it was definitely interesting. It's a good dataset and the overall workflow was quite clear. It was a really great experience.”

Want to participate in an upcoming challenge? View our upcoming competitions on the Xeek website.