Ashwin Jayaraman Email & Phone Number

United States

Greater New York City Area

Teaching assistant for the courses Analysis of Algorithms, Algorithms for Data Science, Databases

Hartford, Connecticut Area

I built a Python Package which allowed the user to create and train entity embedding for the categorical variables present in the data using Keras. A supervised approach was used which allowed different representations of the same variable for different datasets. The package followed a similar architecture to sklearn allowing the user to train their embedding through the fit and transform methods. The final output would be a CSV file which could be used to improve model performance going forward. Embeddings were visualized using the t-SNE algorithm. These allowed us to visualize clusters and group similar values.

My first project at Innovaccer was to implement the PIN, DY and OWR models in R to understand insider trading trend in the historical stocks data. The findings were corroborated from SEBI Litigations using NLP techniques. My next project was to build a document classification model for a law firm in the US. We had a sample of 28,000 documents that were manually classified by a team previously. We first extracted the text from the PDF using the pdfminer package in Python. The text extracted was cleaned by removing stopwords, punctuations, and numbers. We then created a term-document matrix and identified the important words for a particular class using TF-IDF and the chi-square algorithm. These features were then used to train an ensemble model consisting of the SVM and Naive Bayed algorithms. The model was able to identify the class of a document with an F-Score of 0.86 and was appreciated by our customer.Once we moved into the healthcare space, I built a regression model that identified the key drivers of cost at a national and a regionwide level for accountable care organizations. The model was trained upon the quality and utilization metrics with the response variable being the risk-adjusted cost. The coefficients of the variables helped the customer to set high-level objectives for their care manager. I also built and deployed the HEDIS certified measures for our customers. My main role was to identify actionable insights from our customer's data that would lead to a reduction in cost and an improvement in quality. For example, our team created a specialist dashboard that allowed our customer to identify the specialists having a high variance in cost for a similar profile of patients. The proprietary logic to identify specialists was designed by doctors and coded in Python. The final result was then displayed using our in house software and allowed the user to filter and drill down based on location, category, gender etc.

Gurgaon, India

Worked in a team of 3 on a research project for a client of KPMG in the field of Business Intelligence and Analytics and how these are applied in the company to curtail risk,fraud and increase growth.Also researched on upcoming trends such as Internet of Things(IoT) and how Data Analytics can be used with these disruptive technologies to further enhance the performance of business

Pune Area, India

Did an internship in the Actuarial Science department and helped the team in calculating the reserve needed by the company for 2 policies

Master Of Science - Ms, 4.14/4.0, Data Science

Master’S Degree, Mathematics, 9.429

Activities and Societies: Technical Coordinator, Mathematics Society

Bachelor'S Degree, Mathematics, 89.8

Activities and Societies: Executive, Mathematics Society

Education record