Bettina Kozissnik of Data Science Cohort 3 loves the scientific process and the eureka moments that come from hard work. “There comes a point [when] you’re working really hard on understanding [a] concept or getting your experiments to work and suddenly it clicks. That’s a wonderful moment,” she exclaimed. As a biomedical engineer, she pursued a PhD in Oncology to make an impact in cancer research. When she learned that professors spend most of their time writing grant proposals while the grad students, technicians, and Postdocs did the research and worked in the lab, she realized that academia wasn’t for her. So she returned to her business roots and held jobs as an office manager, transportation coordinator, and a health and safety administrator. But she missed research and the eureka moments.
Last summer, Bettina found herself at a data science meetup where she heard local data scientist John Liu speak on machine learning with small data. She was excited about how the field seemed to be at the intersection of her experience in business and science. She spoke with several data analysts and data scientists at the meet up that night and they recommended Nashville Software School (NSS). She attended an info session and made the decision to apply to the Data Science Bootcamp.
The NSS Experience
Bettina was drawn to NSS for the “well designed and thought out” curriculum, the affordability, and the part-time schedule of the Data Science bootcamp. She shares, “I can't tell you, what an incredible relief it is being able to go to school while still working full-time. Yes, it is not easy, but it is much more feasible than quitting your job for a year or two to focus on education for a new career.”
When you work full-time and go to school part-time, there’s a delicate balance between work, family, school, and studying. While Bettina had found that balance for the first six months of the bootcamp, the pandemic quickly threw that off for the last three months. All of a sudden, she had “a little one full-time at home, while trying to work from home and attend school from home.”
She experienced a few in-class challenges when NSS went remote as well. She had to relearn how best to collaborate with her classmates in the virtual environment. “I learned a lot from just seeing code from the other students and having conversations about different coding challenges in my class. Via Zoom that was instantly very different,” she explains. It was also challenging to learn multiple languages in a short period of time. “Python, R, and, PostgreSQL share similarities, but are distinctly different,” she said.
Despite the challenges, Bettina experienced numerous highlights. “Leaving my comfort zone, being able to work with and learn from students with all sorts of professional backgrounds, the "Eureka" moments when the code works, interactions with real life data scientists, learning first hand on the challenges of data science in the real world…” just to name a few.
Enjoy the ride and soak everything up. These 9 months are over much quicker than one would think. Every team project is an opportunity to learn from the challenge itself, your teammates, and an opportunity to also grow personally.”
We spoke to Bettina earlier this year about her mid-course capstone project, Competitive Boston, which takes a look at the competitiveness of the new qualifying times for the Boston Marathon. You can read about her analysis and use her Shiny application here.
Bettina’s final capstone project brings predictive modeling to questions on Stack Overflow, a popular website for programmers where they can crowdsource answers to solve a problem with their code. With so many questions on Stack Overflow, there’s no guarantee that someone will answer your question. Bettina wanted to predict the likelihood of a particular question being answered by the community. Before she could analyze the data, she created a web scraper in Python to extract the questions. After scraping and cleaning the data, she was able to use the natural language processing libraries SpaCy and Gensim to explore the data and find the most common bigrams and trigrams. She also did some topic modeling on the different subsets of questions.
She then built her machine learning model. “By using the LIME text explainer I'm not only able to give a yes or no answer to the question of whether my model "thinks" a specific question would be answered on Stack Overflow, I'm also able to show why my model "thinks" this and which elements of the question tilt the probability in either direction,” she explains. To make it interactive, she created a widget in her Jupyter Notebook that lets users type in their question to see the probability of it being answered on Stack Overflow and why it may or may not receive an answer. You can view her data visualizations from her exploratory data analysis on her R Shiny dashboard or watch her share her findings from her project here.
Bettina hasn’t slowed down since finishing bootcamp. She is currently working through DataCamp courses on machine learning and Python, learning to use Tensorflow, practicing with neural networks, and learning about the services provided by AWS.
She hopes to return to healthcare as a data scientist. While it may not be cancer research, Bettina is eager to make an impact in healthcare again and is looking forward to many more eureka moments.
You can learn more about Bettina on her podcast.
Check out all the recent grads on Data Science Cohort 3’s class website and hear the graduates share their experience at NSS and capstone projects in their podcasts below.