Felicia Anto CHRISTY
AI/Data Analytics Intern
(January 2021 – June 2021)
Over the first half of this year, I had the amazing opportunity of interning at UOB Asset Management (UOB AM). With a presence in nine Asian countries, the company serves individuals, institutions, and corporations, offering a comprehensive range of investment solutions and services. During my time there, I had the privilege of working in the Investment Technology & Support Unit team, who support the various investment management teams, by performing data extraction and analysis tasks and reporting. They also spearhead innovative solutions that use technology for the automation and improvement of existing processes within the company.
Having just finished my first semester in the Masters of IT in Business (MITB) programme’s Artificial Intelligence track, I found myself a novice in the machine learning field. However, when I was offered this internship, I knew that this was an opportunity I could not miss out on. Setting aside my fears and doubts, I decided to embark on this journey, determined to make full use of it, to learn, contribute and grow myself through this experience.
At UOB AM, I was greeted by a team of warm and friendly colleagues, many of whom were a lot higher in seniority. I was humbled to be in the company of such experienced colleagues, and to be able to work closely with them.
The key project that I worked on during my internship aimed to use machine learning methods to improve the Tactical Asset Allocation process, an active portfolio management strategy to tilt the strategic asset allocation, taking advantage of shorter-term market trends and economic conditions. The main component of this project was to build a timeseries prediction model that could forecast asset prices. I worked on this project for the Investment Solutions team, with the support of the Data & Digitalisation team.
Being someone with little to no financial knowledge prior to this internship and little hands-on experience with model building, this project posed as a great challenge for me. The project required me to explore various machine learning models, such as RandomForest, XGBoost, etc. and the use of these models in financial timeseries forecasting. Although daunting at first, with the guidance of my seniors, over time I was finding myself being able to better understand and contribute to the project in meaningful ways.
Throughout the project I have been able to experience first-hand the entire machine learning lifecycle, beginning with data extraction, and pre-processing, followed by feature engineering, model selection and development, model tuning and refinements and finally evaluation of the models. After experimenting with several models and many rounds of data engineering and model refinements, we were able to achieve a series of classification models with satisfactory accuracy scores. Using the predicted prices to perform asset allocation, the back test results were promising with the proposed allocation outperforming the benchmark allocations, in terms of annual returns and sharpe ratio.
Another area I was tasked with was to perform various research works revolving the use of deep learning techniques for financial timeseries forecasting. I read many papers and research that has been done on deep neural networks, such as RNN, LSTM and GAN models, and their specific use case when it comes to financial data. While this gave me the chance to read and grow my knowledge, it also opened my eyes to the importance of research as a Data Scientist. This is an industry where there is vigorous research that is ongoing, and it is vital to keep up to date with papers on new models, techniques, and tools. This is a primary way in which any Data Scientist can add value to the work s/he does.
My colleagues at UOB AM were always keen to share as well as to listen, often engaging me in discussions whereby there was a free exchange of ideas and suggestions. This was very welcoming and encouraging for a young Data Scientist like myself. It was a safe space for me to voice my questions, thoughts, and suggestions which were always followed by active discussions. The team of senior colleagues I worked with were very nurturing and displayed great mentorship, which was very inspiring for me.
While I was doing this internship, I was also attending MITB lessons in parallel. My work at UOB AM and the curriculum at MITB complemented each other, often acting as reinforcements of each other. While my lessons gave me in depth understanding into many machine learning and deep learning concepts, my tasks at UOB AM gave me the chance to apply these same concepts into practice. The modules I found especially relevant and helpful were Introduction to Artificial Intelligence and Applied Machine Learning. These modules introduced me to the fundamental concepts and algorithms in machine learning, which were the basis on which I was able to perform my tasks at my internship.
Overall, my time at UOB AM was a challenging, yet equally rewarding experience. My greatest takeaways would be the many conversations and discussions that I engaged in with my seniors, which have been pivotal in helping me to perform my tasks to the best of my abilities. The environment as well as the project I worked on have helped me to learn and grow tremendously, and the lessons I have learnt are ones I will carry forward with me on my Data Science journey.
I extend my sincere gratitude towards my colleagues at UOM AM, for their thoughtful guidance and great mentorship, specifically to Jeff, Ryan, Andrew and Catalin, with whom I have worked closely during this internship. I also extend my thanks to the SMU MITB programme, for helping me in securing this opportunity and for their constant support and counsel throughout the internship.
