My interviewer was a Lead Data scientist at Emids and the interview lasted for about 30 minutes. Initially, questions were about my profile, my previous background, and past work experience. Then, he gave me a case study about state election result prediction and asked about how I would approach the case study and what should be the sampling technique, how to build a model, and what should be the parameters for the evaluation, etc. Then, the interview took a slightly technical turn and questions were asked about the interpretation of the evaluation criteria, Type I error and Type II errors, then he gave me some time to prepare a small story around the case study and asked to present the technical results in layman’s terms.
3 Months of Continuous Learning :
With COVID-19 pandemic devastating the entire world, Emids was helping the U.S. healthcare payers and providers in the digital transformation process. I was lucky enough to intern at Emids during such critical times, where I worked on two predictive analytics use cases.
The first one was about the optimization of hospital infrastructure planning by building a machine learning model that can accurately predict the length of stay (LOS) of patients at the hospital before admission. With the help of accurate LOS prediction, hospitals will be able to better plan their staff, bed allocation, and doctor visits. Also, this helps insurance companies (payers) to have a better idea about the amount that a patient might claim for the treatment.
With people struck at homes due to lockdowns for a prolonged period of time, mental health was a major concern so we tried to address this problem in another use case by predicting the risk of severe mental illness based on multiple factors.
Data for both use cases were in raw format and couldn’t directly be used to build the model. So first, we cleaned the data, imputed the data for the missing values, then we carried out the exploratory data analysis to find out the hidden insights in the data. Feature engineering was done on the cleaned data. Multiple models were built on the data to check which algorithm would better predict the results. Models were later improved using hyperparameter tuning techniques. In the end, use cases were presented to higher management and they were satisfied with the use cases.
I’m very thankful to Emids for honouring the internship offer even during such difficult times, the entire onboarding process was very seamless. My mentor helped me a lot during my stint at Emids, under his guidance, I could complete multiple machine learning certifications. Also, my manager was very kind and supportive, despite his busy schedule, he made time for us.
In simple words, my internship was a great learning experience!
Comments