Interactions with entities make the world go round! When a person is about to interact with various entities, at various capacities, donning different hats over shorter, longer lengths of time he needs to know the "reputations"​ of these entities ahead of time to avoid being harmed by them. When we navigate internet: are certain URLs safe? When we go to a restaurant: How good is it? How safe is the locality?When we go on a pleasure trip: How good are the transportation facility? What is the crime-rate of the surrounding area? Are the roads navigable? Are there communal tensions prevailing in the area? Can we forewarn or intimate a person ahead of time using Internet Of Things (through personal wearables)about an imminent interaction with "entities"​? Again, depending on the context, intent of his interaction and personal profile the "reputation"​ of the interaction will vary. This research involved the following salient aspects from a Data Scientist's perspective----developing Instance based Regression model (a type of Lazy learner) on count data of incidents to determine reputations of locations----providing means of incorporating domain knowledge(e.g. from criminology, behavioral studies etc) in aforesaid model----developing negative Binomial based regression models as alternative for the above----Fusing information of disparate type obtained from disparate sources by means of Natural language Processing: Web crawling, Government database etc.On to a different project: Providing an interface which interacts with human beings in natural language(English)is a dream for everyone. For example, a network administrator has to negotiate so many drop-down menus, edit-boxes and other controls to arrive at the suitable appliance or product ----By analyzing short description (or query) entered in natural language semantically show the relevant capabilities of the appliances which will most likely address the problem of the end-user(say, network administrator).

Economist Ronald Coase most probably had said “If you torture the data long enough. nature will confess.” Somewhere along its circulation it has been distorted to “If you torture the data long enough, they will confess.”. I like playing the role of torturer in the former sense to uncover the truth hidden in the data!!!Impetus has entrusted me with the task of extracting these "confessions" from the data and devising automated means of "torturing" data-Big-Medium-Small.--- Developing predictive models on click-stream data augmented with demographic and psychographic information. Mathematical models used were: Multinomial Logistic Regression, Random Forest; Continuous Time Markov Chain Models---Sentiment Analysis on clients e-mail, communications etc. For topic modelling Latent Dirichlet Allocation and for information retrieval and analysis variants of Latent Semantic Analysis were used

---Developed algorithm and implemented Model based and data based methods for fault detection and diagnosis (FDI) of electro-mechanical systems--e.g. In Aerospace Domain, FDI of aircraft engines and auxiliary power units; in Automation and Control Domain, Equipment Health Monitoring by vibration and other signal analysis.---Used Natural Language Processing(NLP) based methods for processing Field Service Reports to build a case base for maintenance of aircrafts--- Developing and implementing Rule based Fault Detection and Diagnosis methods for Heating, Ventilation and Air Conditioning (HVAC) systems --- Developing and implementing mathematical models involving ordinary and partial differential equations for gas separation processes by membranes for various configurations and geometries

Development of algorithm and implementation of Classifiers for document classification system. Dealing with Documents containing short descriptions and scaling up for processing humongous volume of data were major challenges

Developing mathematical models for process industries.Developing algorithms and implementing Fuzzy-logic based Advanced Control System for process industries