UQ masters student secures data science job post Unearthed Brisbane win
Along with team member Leopold Fournier, University of Queensland Masters student Arun Prakash won second prize at energy resources hackathon Unearthed Brisbane 2017, which took place over the 21-23 April weekend at River City Labs and the Startup Precinct.
Team Red developed their web application in response to Origin Energy's "Well of the Future" challenge that asked participants to find ways to determine the pressure or level in a well without a gauge. PFinder uses machine learning algorithms to create a model that estimates the downhole pressure at the bottom of each well, in real time and with a good accuracy. Consequently, it does the work of a gauge on wells that don’t have one. It also feeds on the data provided by gauges to improve its performance.
On the second day of the hackathon, Arun and his fellow participants heard from Dr Penny Stewart, Managing Director of PETRA Data Science about her experience participating at previous Unearthed events and her role in delivering engineered data science solutions to major mine sites. Following the talk, Arun sent Penny a LinkedIn request and she later responded congratulating him for his second place win. She also mentioned that she was looking for a data scientist to join the team at PETRA Data Science. Arun applied for the job and was subsequently hired after an interview at their offices.
Dr Penny Stewart said that Arun is already proving to be an excellent member of the team.
"He is currently working on a machine learning project to automate conveyor drives warnings, as well as a big data study to identify how the best truck and shovel operators achieve their outstanding performance," she said.
"Securing this role would not have been possible without attending Unearthed Brisbane. Thank you to the Unearthed team for organising such a great hackathon" - Arun Prakash.
Considering his recent success, we asked Arun to share his thoughts on how best to prepare for a hackathon and increase your chances of a win.
What do you think makes for a successful hackathon team?
Any interdisciplinary team with 2-3 people is a great team. I recommend a 2-3 person team, as Unearthed hackathons are only 54 hours long. If there are more than 3 people in the team, time may be wasted on discussions and agreeing or disagreeing on ideas. If it is a data hackathon, make sure your team has a machine learning expert, full-stack web developer, and someone with business analysis or industry/domain knowledge.
Which skills did you bring to your team and how did you form your team?
There were only two people on our team. I have experience working as a software developer as well as being a full-stack web developer. Now, I am pursuing my masters degree in computer science at UQ, focusing on machine learning and artificial intelligence. Therefore, I applied both my machine learning and full-stack web development skills for this project. My team mate Leo focused more on the business analysis side.
Why did you attend the hackathon and what did you get out of it?
I have been doing side projects in machine learning for a long time. I came to know about Unearthed Brisbane through the UQ computing society. I immediately went through the hackathon web page and the challenge details on the Unearthed portal and it really interested me. So, I thought it was the right time to participate and test my machine learning and web development skills in solving real industry problems.
It was a rewarding experience. This hackathon showed me the importance of machine learning in the mining and energy industries. There are many more problems that need to be solved using machine learning in the energy sector.
What would be your top three tips for people participating?
Great question. Here are my top three tips:
1) Time management is key:
Unearthed hackathons are 54 hours long. So, you have to optimise your time really well. For example, if you are planning to use Amazon Web Services during the hackathon, try to spend some time on familiarising yourself with AWS before attending the event. In my case, I planned to use python programming language for the machine learning part. So I practised end to end machine learning workflow before attending the event. Also, take a printout of some useful programming cheat sheets. Here are some that I have collected and published on my blog. Cheat sheets can save time during programming.
2) Understand the problem description really, really well:
Make sure you understand the problem and all the edge cases before you start writing the code. My team mate Leo and I discussed our understanding about the problem with the Origin mentors and ensured that we were on the right track. I would advise finishing all of your important discussions about understanding the problem before Saturday morning. That way, you can focus more on solving the problem on Saturday and Sunday.
It is best to solve the problem AND be able to sell it to the judges. Make the judges’ job easy. It is ideal to implement your machine learning models as a web application. Make it easy to use and easy to integrate your work into their existing system. Pitching and presentations scare me, so I attended the practice pitch session organised by Unearthed team. It was really helpful feedback for the final pitching event.
What do you think are the main barriers for people attending hackathons?
I was aware of Unearthed hackathons in 2016, but I didn’t attend because I thought I wasn't skilled enough to participate. I think many potential participants feel the same way. So, it would be good to provide useful resources to acquire those skills. Also, publishing hackathon winners and participant’s interviews would help in allowing future participants to learn from previous participants' hackathon experience.
Has attending the hackathon event changed the way you think about innovation, startups, and entrepreneurship in the resources sector?
Yes, Unearthed hackathons really helped me to understand how startups work and how to build a great prototype in a short, focused time. Also, it is really encouraging to see that machine learning can solve many problems and increase revenue and productivity in the mining and energy industries.
Anything else you would like to let people know about your experience?
If it is a data hackathon, I would recommend using Python programming language, as it saves a lot of time when you build a web application for your machine learning model. Also, prepare a rough tech stack and practice/collect useful resources before attending the event. For example, here is my tech stack I intended to follow for this hackathon:
IDE: Jupyter Notebook, Atom
Data Cleaning/Exploratory Data Analysis/Visualisation: Pandas, Numpy, Seaborn, Bokeh libraries
Machine Learning: Scikit-learn, Scipy, Tensorflow, Keras libraries
Finally, in my experience, what judges care about most is:
1) Whether your solution can actually solve their problem
2) Whether your solution can increase the company's revenue and productivity
3) How easy it is to integrate your solution into their existing systems
Unearthed would like to once again congratulate Team Red on their success at Brisbane this year, and a big thank you must go to Arun Prakash for taking the time to share the insights he gained from participating.
To register your interest for our upcoming Digital Tribes hackathon, visit: https://www.unearthed.solutions/digital-tribes-perth/
For more information about upcoming Unearthed events, visit: https://portal.unearthed.solutions/