Internship Experience at DSAID
Hello! I am Hari, a second-year student majoring in Data Science and Analytics at NUS.
Over the past few months, I’ve been interning in the Quantitative Strategy team, which is part of the Data Science and Artificial Intelligence Division (DSAID) in GovTech. I am with a forward deployed team at the Ministry of Manpower (read more about DSAID’s forward deployed teams here), and my main task for this internship is to work on a Natural Language Processing (NLP) project.
Have you heard of MyCareersFuture? It is a job portal jointly developed by Workforce Singapore and GovTech to help Singaporeans find relevant jobs based on their skills and competencies. My project focused on developing a machine learning model to predict the correct occupation (according to the Singapore Standard Occupational Classification) using only the job title and the job description from the job advertisement on MyCareersFuture. There were 2 parts to my work over the course of 12 weeks: the first 6 weeks were geared towards data engineering while the last 6 weeks was more data science focused.
Data Engineering
Data engineering was completely new to me, as I hadn’t learnt anything about it at school. I was tasked with constructing an automated ETL (Extract, Transform, Load) pipeline for ingesting the job advertisements data. Over the weeks, I managed to learn the basics of AWS (Amazon Web Services) and was exposed to the different products that AWS had for processing and storing data on the cloud. I also went through AWS’s Cloud Practitioner course to pick up the skills needed to implement the data pipeline. Shaun, the team lead, was kind enough to constantly check if I was interested and okay with the work I was tasked with, which made me feel more assured and comfortable while working on it. We adopted a weekly sprint practice where he would task me something new every week. It was daunting at first, but everyone was willing to lend me a helping hand when I was stuck, and I was encouraged to ask as many questions as possible. In addition, I learnt to work well together in a tech team with different roles — for example, I picked up good Git practices, was exposed to the Google Standard of Code, and developed unit tests for the first time.
Data Science
When it came to the data science portion of the work, I was relieved, thinking “finally, my area of comfort!” However, I was challenged with the task of researching on how Bidirectional Encoder Representations from Transformers (BERT) was trained, which led me to read up on masked language modelling (MLM) and next sentence prediction (NSP). Most online resources are more geared towards fine tuning the final layer of BERT rather than updating the original weights themselves. To learn how BERT was trained, I read many research papers and code repositories, and spent time discussing ideas with Shaun and my colleague Ben. Through this process, I gained insight on the history of NLP models and the reason BERT outperforms them. Armed with a better understanding of MLM and NSP, I put my knowledge to the test by writing a production-ready machine learning script according to industry practices so it can be deployed to train other language models. This is still an ongoing process but I am happy to say that I am close to completion!
DevOps
I was also exposed to some DevOps work as Shaun and Ben were working on deploying the trained model on Container Stack, a GovTech-developed tool for deploying applications on the cloud. Through this, I was exposed to Docker and Kubernetes and was able to understand the importance of these tools, especially when it comes to the deployment of our data science work. It sparked an interest within me to learn more about DevOps, and I will be actively working on this over the next few months.
Fun and Games
Working in GovTech is not just all work. We made sure to make time for fun too! In addition to friendly and helpful colleagues, there is a games room in the office which I often visit with my colleagues especially during Friday lunch and after work. We often play air hockey and foosball, although Shaun always beats me at foosball and I have yet to win a game against him. We also went out for a cohesion activity where we did art jamming. Trust me when I say I am not artistic — my painting was just shades of blue. We had a good laugh about that. It was refreshing to see the creativity and artistic talent of my DSAID colleagues come through in their paintings, and fun to hang out with everyone on the team.
Overall, GovTech is a great place where I have the opportunity to learn and develop my technical skills while also building good relationships with my colleagues. It is definitely a place I would want to work at as I have seen how it is possible to have work-life balance through my own eyes. This internship has surpassed my expectations and changed my perception of GovTech. I would like to come back and work at GovTech after I graduate. With nice colleagues, challenging work and games. What more can I ask for?
If you are keen to apply, visit go.gov.sg/govtechinternship to find out more. Curious to learn more about DSAID? Visit https://medium.com/dsaid-govtech to get to know more about the work they do! All the best and I hope to see you around!
