Soumyadipta Sarkar

I work as a senior member of the North America catastrophe modeling team. We develop catastrophe models for various perils, generate hazard scores, relevant statistics, and create analytical products for our clients. This involves developing efficient algorithms for processing very large amounts of GIS and census data with Machine Learning and other statistical tools for creating risk metrics.

- Lead a team of analysts to investigate and validate different catastrophe risk models. - Work closely with research and development teams and numerically validate Hazard, Vulnerability and Financial modules of AIR catastrophe models. - Management and manipulation of large exposure and model data with R and SQL. - Develop and maintain automated analytical validation reports using R and Python. - Help clients to make data-informed decisions and provide support.

My work focuses on developing efficient algorithms to analyze the behaviour of large scale vibrating systems under random excitation such as earthquake. I investigated the choice of dominant modes and developed a hybrid algorithm for response analysis of such systems. For this, the random forces and material properties are modelled as random processes in a stochastic finite element framework. Further, I investigated sequential and parallel implementation of these algorithms.

My work focuses on building artificial neural network (ANN) models that can be used to estimate earthquake intensity. I used fault and site data from Kyoshin net (K-NET) database from Japan to train the model to obtain accurate predictions of peak ground acceleration (PGA) at site.