Architect Anupama Garla fell in love with data science and dreamed of combining the two fields. After Anupama discovered that Metis operates a Live Online Data Science Bootcamp, Anu followed a friend’s advice and chose to apply! After three months of data science projects, learning Python and Spark, Anu is now on the job search. Anu describes the online classroom and learning style at Metis, her advice for other career-changers, and her plans to innovate the world of architecture now as a data scientist.
What inspired you to transition from architect to data scientist?
I was working on a Ph.D. in Urbanism and Societal Change, but data science had been on my radar for a while. I had done a lot of data visualization and analysis, research-driven and research-based projects in my architecture career. I took a pause on the Ph.D. program to become a Business Analyst and even though I excelled at the Business Analyst role, it became clear that it wouldn't help me to transition to a job in data science, so I stopped after six months. I took that time to focus on self-education, so I took some Python and SQL programming courses with Coursera. After listening to the podcast OC Devel’s Guide to Machine Learning, I then took the courses he recommended. All of the information I learned was a little overwhelming. There were so many tools, and I didn't know how to put it all together. I knew a data science bootcamp was the next step.
Why did you choose Metis’s Live Online Data Science Bootcamp?
A friend of mine was a graduate of Metis, and he was really happy with it. I liked that the bootcamp was remote. The on-campus bootcamps around me would have been a long commute, and when I heard that bootcamps can take up 80 hours of your week, I didn't want to spend two hours a day in the car.
What was the admission and interview process at Metis like? Did you have to complete any pre-work after you were admitted to the bootcamp?
Metis gave me a coding challenge with a little data analysis at the end. My interview was with a Metis graduate and we talked about where her career went after graduating. She also came from an architecture background! She gave me brain teasers during the interview. I didn't need to know the answer to them; she wanted to observe the way I think. Once I was accepted, there was required pre-work to complete before the first day of class.
Since you took the remote data science bootcamp, what did a typical day look like for you?
The bootcamp is live online, and runs on Central time. Every day ran from 9am to 5pm Central time, which for me was 7am to 3pm. The schedule worked really well for me. We did an hour of pair programming, and then a lecture for two hours. After lunch, each student took a turn presenting a technical talk on something related to Data Science. I did mine on mapping and visualizations, but I thought the best one was a colleague’s talk on bias. We then had our standup and discussed our projects. In addition to project presentations, there was an hour where the teacher would be online and students could discuss any roadblocks or receive help from them. You could also book one-on-one time in advance but I didn’t do this very often. For the last month of the bootcamp, we worked on our final project.
What did you learn in the Metis live online curriculum?
We learned Python, SQL, Nano, and HTML, and we used a bunch of packages, like scikit-learn, Numpy, Selenium, Beautiful Soup, and Geopandas. People also used whatever was specific to their interests. I used Google Cloud Platform and AWS to host data and run processing. We learned about big data tools, like Spark and Dask, and recommendation systems, NLP, and topic modeling methods using NMF, LDA, and LSA. Finally, we learned about neural nets.
Did the teaching style match your learning style?
There were three different teachers and each one taught their strongest subject. I favored one of the teachers because her slides were clean and to the point. She was visual in her way of teaching, and her explanations were simple. You are required to do a presentation at the end of each project you completed, and to simulate the work environment, you are graded on how well you explain the project to non-technical people. I have experience with this from my background in architecture, so I excelled at these presentations.
How did you interact with your instructors and fellow classmates in a remote bootcamp setting?
My cohort was tighter than I had imagined we would be in an online format. We were remotely connected to Metis’s in-person Chicago class, too. Both cohorts were viewing the lecture as it was happening on location and could ask Slack questions or speak. At first, it was strange to be an online addition to an on-campus class, but in the end, I was happy that they did it like that. It wasn't just a bunch of people online; we were all part of a real classroom.
Were you able to do group projects in the remote Metis bootcamp?
Our first project was a group project! I'm not normally a big fan of group projects, especially for the first weeks of a remote program when we barely know each other or the tools, but it did bring us all closer. The project was built around a marketing agency that wanted to reach out to subway riders and get them to sign up for a Women in Tech mailing list. This marketing agency was also advertising a gala event. We analyzed MTA turnstile data based on day of week and time of day, which we used to distinguish commuter versus tourist traffic as well as pinpoint the high volume times and flow directions to target. Initial analysis of the top 5 commuter stations based on volume yielded a concentration of metro stations in midtown NYC, where the density of office buildings and technology companies is at its highest. However, we decided to analyze clusters of geographical areas to weight our analysis towards a diverse ridership using a k-means algorithm to identify the clusters. We responded to the fictional client with the option of choosing to send street teams to the top 15 commuter stations overall, or diversifying the pool of stations by choosing the top 5 commuter stations in each borough cluster.
What was your final project at Metis?
My last project linked back to my architecture career, and centered around Los Angeles homeowners and the new legislation that streamlines backyard houses. I built “YIMBYme,” a tool L.A. homeowners can use to predict the potential income from building a backyard house. I used a transfer learning methodology where I trained an XGBooost Regressor model on an Airbnb L.A. dataset (augmented with geo-spatial L.A. city data). I then applied that model to L.A. residential lot data to produce income predictions for each eligible L.A. lot. I also segmented the Backyard House Database into income tiers targeted for specific backyard house-building startups and visualized the segments on a map of L.A. using Tableau. I think it's a proof of concept of how Data Science can inform new practices in architecture, urban design, and real estate. I wrote a Medium blog that goes deeper into it!
How did the COVID-19 pandemic affect your online learning with Metis?
Everyone went online due to the pandemic, but it was a natural transition for me. The difficulty came when schools were shut down and my kids were home all the time. I had a lot of help and support with the kids from my parents and from my husband, which was crucial. It is possible to do bootcamp while kids are at home, but it’s more challenging.
What is your advice for setting up a remote learning space, especially when you have kids at home?
My advice to anyone who is taking a remote bootcamp while having kids at home is to streamline your process and be very clear about your goals. As for the actual space, it's good to have two screens, one for the lecture and one to follow along. Your computer should be able to run the packages you are required to download, so be sure your computer isn’t too old! The reason I used the Google Cloud Platform is because I couldn't use MongoDB or Docker on my computer, and my computer is from 2016. It’s not even that old, but it wasn't good enough!
How has Metis prepared you for the job hunt?
During the first three months of the bootcamp, we learned about the tools for a job hunt, like refining your LinkedIn profile, the importance of networking, and developing your blog and github. There were also guest speakers who would do career lectures. They talked about their work and would announce if their companies were hiring or not. At the end of the bootcamp, you do your final presentation for Metis’s hiring partners. Metis also sends out your resume and presentation video to their hiring partners. I had one company reach out to me after my final presentation. Since graduation, I meet with a Metis career coach once a week. Metis also offers technical interview practice, but I haven't taken advantage of that quite yet.
What has your job search been like so far?
COVID-19 has changed the plans I had. The companies I was interested in – like Uber and Airbnb – are location-based and they are not hiring right now. I'm open to Data Scientist, Data Analyst, or any data science position that is more about making a business case than narrow technical work. I've been looking at LinkedIn and Built in LA, and I'm on TechCrunch’s mailing list to keep track of which companies are being funded. Regardless, there are still a decent number of jobs being posted. My husband now works remotely, and if we both work remotely, we can live anywhere!
Did attending an online bootcamp help you prepare to work remotely?
Yes, definitely. I'm still sitting at the same desk in the same room. I was completely ready to be a remote worker after learning remotely with Metis. I am also good at self-discipline and doing what I need to do, so that helps too.
How are you continuing to learn data science after graduating from Metis?
I'm still working on a project I started at Metis. It's a children’s illustrated book recommendation system that I want to develop into a Flask app. That will take some big data tools because the entire data set will be too big for my computer. I have also been attending virtual meetups; one about recommender systems and a AI for Mankind COVID meetup that introduced projects to collaborate on. Recently, I went to the WiDS - Women in Data Science Conference Online, which was a phenomenal experience, so I am starting a WiDS chapter in L.A. I’ve been looking into attending more meetups, but I'm really driven by developing my own projects right now. I've started reaching out and making myself known to the architecture and real estate world about my project and how my unique knowledge could be useful. My brother and I are also toying around with developing a business around backyard houses and exploring data science tools that could uniquely benefit such a business.
Are you happy that you went down this route into Data Science? Do you think that right now is a good time to get into Data Science?
Yes I am, although I wish I had done Metis a year ago before the COVID-19 pandemic happened. It's hard to change your career and know the right timing to do so. But a lasting effect of COVID-19 is that I think that a lot of companies are more open to working remotely now, which is good. And there are so many online resources for those looking to get into tech, including people willing to help you out.
How does architecture inform your data science career?
They are very similar professions in terms of their processes and how they bring value to people. Architecture is a technical role; it's your job to make it relevant to your client and help them to make decisions. It's a consulting position on a technical matter, which I think is essentially a major part of what Data Science is. The process of architectural design is to gather information in visual and physical form, process that into a design, and evaluate that design to make sure that it supports the goals of the client. The process of data science is gathering data, making models, and evaluating the metrics to give the client the desired results. It's the same process but using completely different tools. I'm good at visualization and making presentations that tell a good story, and I think this will carry me far as a Data Scientist. The architecture I engaged in was very research-based. We looked at history and aimed to predict the future of certain industries, cities, or cultural phenomena. Only after understanding the trends, did we intervene with our predictions, using architecture, urban design, or research projects.
What was your biggest challenge as a career changer?
Besides time and money, I think it was psychological. Most people don't think that architecture has anything to do with data science. They thought I should do graphic or UX design instead. People don't know what architects do, and it's hard to tell from the outside unless you know one, especially my type of architecture. I had to have faith in myself that going into data science was the best choice for me.