The Co-op Close-up: Data Science at ICBC

SFU’s professional master’s program in computer science trains computational specialists in big data and visual computing. All students complete a paid co-op work placement as part of their degree. In this feature, we examine the co-op experiences of some of our big data students.

Sagar Parikh completed his bachelor’s degree in information technology at Gujarat Technological University, India, before joining the professional master’s program at SFU. During his undergraduate studies, he completed internships working with Java, Python, cloud services, and graphic design.

Can you tell us about ICBC? What is it like working there?

The Insurance Corporation of British Columbia (ICBC) has been the sole provider of mandatory, basic auto insurance in the Canadian province of British Columbia since its creation in 1973. ICBC also sells optional insurance and provides driver licensing and vehicle registration and licensing. The crown corporation has multiple offices across BC with its head office in North Vancouver. I am working as a part of the strategic analytics team, which falls under the finance division. My primary work is on historical claims data at ICBC.

ICBC is committed to creating a workplace where employees know they are valued and where employee experience is a priority. Its core values are at the heart of the organization, and ICBC’s employee commitment guides it in promoting growth, well-being, and a culture of caring — for the customers and the community.

Can you tell us a bit about the project(s) you are working on in your co-op position?

One of the main responsibilities of the data science team is partnering with ICBC’s claims division to improve the claim handling and settlement process. ICBC receives more than 600,000 claims every year, with more than 10% of these having an injury component. One way analytics can help is by identifying claims requiring more attention, and thus ensuring fair settlement. The information required to do so is not always present in a structured format in the relational databases, but oftentimes it is present in the notes — unstructured data — taken by staff.

One of my main tasks during this internship is to establish a benchmark, or proof of concept, for utilizing NLP techniques to extract useful information from this huge untapped source along with generating various predictive models. Such models would be able to assist in real-time analysis and identification of claims requiring more attention. The approach was to start simple, with bag-of-words methods, to establish benchmarks of performance, and later move to more complex techniques, such as word embeddings and language models.

How did the big data program prepare you for your co-op?

While working on multiple projects here at ICBC, I have interacted with and used multiple big data and machine learning tools and technologies such as Pyspark, Spark NLP, Hadoop, Sklearn, Elephas, Keras, and so on. The applied knowledge gained from the big data courses at SFU combined with my background in computer science and my interest for data science came in more than handy during this work term. Working with big data tools such as Hadoop and Spark was especially smooth because I worked with them previously in my big data lab courses. These courses were particularly useful to me for this internship since I’m working on huge volumes of data fairly regularly. The datasets that I work on sometimes have anywhere from 3 to 30 million rows! The assignments and project work in the big data lab courses prepared me to handle such huge quantities of data and cluster resources efficiently.

What are your most valuable takeaways from this co-op experience?

The highlight of the work term has been the opportunity for constant learning and feedback. Positive feedback and appreciation have helped me alter my path whenever needed. In terms of technical skills, the best takeaway for me was the practical NLP knowledge I gained during this term. I learned a lot and am still learning about the overall flow of NLP projects, and the intersection of machine learning and deep learning with NLP. Armed with this knowledge, I am now more confident than before in pursuing a career in data science, especially in the NLP domain.

Do you have any advice for other co-op students?

The ability to quickly absorb and adapt to the transition from coursework to co-op played a part in my success. The key to this ability I believe is effective communication. A lot of my success has been due to the constant check-ins and bi-weekly meetings with my supervisor and manager. They answer my questions patiently and point me in the right direction when I’m stuck. So don’t be afraid to ask questions.

The role of a data scientist requires one to truly understand the data before generating insights or building predictive models. In my experience, understanding the business is the first step towards understanding the data, so my advice is to try to learn as much as possible about the company you are working for, and in particular, understand the data flow within the organization.

What do employers say about our students?

“Sagar’s most impressive strength is his initiative. He is confident and resourceful enough to suggest modifications to assigned instructions when they would benefit the project, and is able to independently do all necessary research and review of documentation required to achieve his assigned objectives. He is very easy to instruct, quickly absorbs new concepts, and efficiently requests assistance and clarification when needed.”

— David Menard, Manager, Data Science, ICBC