Welcome to the Grakn Labs team — Oscar Darwin
This summer we have two interns working in our office, learning all about Grakn and investigating various applications of and integrations with our distributed knowledge base.
We’re excited to have them as part of the team, and so we figure we’d revisit our ‘meet the team’ series, and find out what their background are and what they’re working on with us on Holloway Road.
This week, hot on the heels of his breakout post on how to use Grakn for rules-based machine learning, we’ll introduce you to Oscar. Let’s get to know the man behind the code!
Hi, I’m Oscar and I’m passionate about technology and mathematics.
So what do you do at Grakn Labs? What makes you tick about the technology?
I’m a machine learning intern, so my job is to use Grakn in new and exciting ways! I’ve been at Grakn for 2.28 weeks so far. The job is challenging because there are so many possibilities of machine learning algorithms and data sources to choose from that picking the best one can be difficult! It’s also rewarding as I get to read about all these new advances in the ML research space!
I’ve been loving working in this space and actually wrote a paper about the travelling salesman problem last summer (a well known problem in computational graph theory). The paper is published here. In general, the idea of knowledge bases is really exciting as their representational power and also understandability is so much greater than typical relational database approaches. In 25 years, I’d love to see the knowledge gap between those in tech and those not be bridged — and technology like this is a great way to achieve that.
What did you do before working here?
I’ve been studying maths and computer science at Oxford and just finished my 3rd year where I particularly enjoyed learning about the mathematics behind Brownian motion (which is the model used to describe the movement of pollen particles in water). Prior to Uni, I designed a machine learning system for a mobile advertising company where we modelled successive clicks on mobile ads (by the same users) and built a classifier to determine if clicks were fraudulent. In hindsight, Grakn would have been incredibly useful for that company’s data structure!
What else do you have planned for us during your internship?
For the rest of my internship I am looking at bridging the gap between knowledge bases and conventional machine learning techniques by building a connector to TensorFlow. Essentially, this involves converting from Grakn’s graph-based model to a set of numbers describing the knowledge base. Following this, I am hoping to apply TensorFlow to some kind of dataset which has a lot of graphical structure and real world applicability. If you have any suggestions HMU!
Aside from Grakn and maths, what else are you into?
In my spare time, I do a lot of singing and classical music composition. I play the piano and sing in an a cappella group. Recently we made it onto the choir show “Pitch Battle” with guest judge Joe Jonas! I think the office got a kick out of seeing me dressed up and made up on the show. In August, we are going up to perform in Edinburgh fringe, so every morning you’ll see me on the Mile. Come say hi if you’re in town!
(Editor’s Note: you can see Oscar’s performance on Pitch Battle here.)
