This summer Andrea Rosendahl, BSc Geophysics at the University of Bristol, and Axel Øvrebø Harstad, MSc in Computer Science with AI as specialization at NTNU, worked as EMerald Geomodellings first ever summer interns.
Different studies with common ground
Though different studies, the common ground that made them sought out a start-up was the opportunity to solve real problems in an agile workplace. “I started looking for internship opportunities once I was ‘evacuated’ from my exchange year at UC Berkeley in the middle of March. As an AI student I was primarily looking for an internship where I got to learn, apply and experiment within the field of machine learning. I found Emerald Geomodelling, which could offer just that.”, says Axel.
Andrea was looking through a list of different companies where geophysicists were working and found Emerald Geomodelling. “When I read more about the company, the people and what they were doing, I decided to apply for an internship.”, she says. “It was interesting how they are using a unique method in retrieving geophysical data. During my studies, some of my favorite modules involved geophysical surveying, and I wanted to see how it is applied in real life.”
The code of covid-19-summer
While corona made most of the world stop this summer, Andrea worked with forward modelling. “Forward modelling is basically using a model in order to create an output. I used the inversion program AarhusInv to apply data to create models about the subsurface. We used programming to execute the code and created resistivity models.” Andrea explains.
Axel edited the script in python. “This made the code possible to use for our problem-solving.”, she says. “I created an idealised and simplified subsurface which I used as input in the python code and the inversion program. The outputs from this program gave resistivity models based on the initiated subsurface parameters." says Andrea
Partners in code
In order to extract the models, Andrea relied on new machine learning algorithms. “I got to explore and adapt machine learning algorithms for modelling bedrock topography, more specifically estimating the depth to bedrock." says Axel. He and Andrea worked together to make codes adapted to the forward modelling and setting up an interface for testing synthetic models.
“The goal was to see how the program works and identify strengths and weaknesses. This is critical information using the program further, because then we could compare the “real” to the inverted subsurface (output from inversion program). This would make it possible to compare the results and identify where it shows an accurate representation of the subsurface and where results are more questionable.” says Andrea.
“A very interesting new approach that we explored was to view the resistivity profiles as 2D data, in contrast to individual profiles (1D), and use convolutional neural networks to predict the depth to bedrock at each resistivity profile. This approach is intuitive in the sense that the model will take neighboring profiles into account, when estimating the depth to bedrock at a specific profile. Also, we tried converting the problem from a regression problem predicting the exact depth in addition to uncertainty, in order to solve a classification problem, where depths were divided into different classes. This made it possible for the model to predict numerous intervals in which it believes the depth belongs to, each with a corresponding probability.” Axel explains.
The Emerald family
According to both Andrea and Axel, a summer internship in a startup is very interesting and exciting. “Everyone has been very open and welcoming. It was nice working in a small group where everyone knows each other, teamwork is in the core and nobody says no to help or to contribute.” Andrea says. She continues “After the summer I feel that my work is appreciated and made a difference in their further development, making the EMerald solution even better. The search for continuous improvement is very inspiring – especially since that is what you do as a student. Also, it was motivating to see how much effort they put into the work to get results back to the clients, and their willingness to work hard in order make the company grow.”
“My internship at Emerald was an extremely interesting, educational and fun experience. In addition to working in a competent and engaging environment, everyone was very open, nice, and eager to help. I felt included in the Emerald family from day one. It was very cool to see how interested everyone was in what I was working on. I really feel like my work was appreciated, and that it inspired their further development.” Axel says.
“I was given a lot of freedom and flexibility in the way that I worked and with regards to what approached I explored. However, I worked closely together with the other employees at EMerald, especially Arnaud, Craig and Egil. We had frequent meetings, where I would present ongoing results and discuss future possibilities.”