- Worked collaboratively with various clients to develop strategy and approach to defining business challenges to be answered by data science- Provided new insights of the employee recognition data from HR and deployed a renovated evaluation model by adopting network algorithms to recommend employees recognized by peers. The model increased subjectivity on the evaluation process as well as improving automation capability.- Built up an industrial standard model for a security vulnerability scoring system out of predicted mitigation factor analysis using XGBoost machine learning algorithm and exponential decay statistical model.(Patent Pending) - Analyzed tech-talent attrition/retention data and delivered deeper insights of trends and causes depending on physical locations, seniority and tenures of employees to leaderships.- Using graph algorithms to detect communities involved in historical ghost broker data or any candidates data with Neo4j, a graph database management system, and the model is expanding its domain to detect different types of insurance fraud cases.- Provided a working model for anomaly detection of applications and notified automated alarms judging by implemented time series forecasting techniques. The model was running in real time as the system was pipelined with Hadoop and Kafka.

– Designed and executed data analysis of prevalence, risk factors and outcomes of medical records of patients by implementing Dynamiclogic algorithm in Machine Learning (python)– Performed data architecture by manipulating a large number of features to select/create better predictors for enhancing the model’s performance and structures them to be used as an input data for the analysis– Implemented data analysis algorithm in Big Data analysis Environment with Spark

– Designed and executed data analysis of prevalence, risk factors and outcomes of medical records of patients by implementing Dynamic logic algorithm in Machine Learning (python)– Performed data architecture by manipulating a large number of features to select/create better predictors for enhancing the model’s performance and structures them to be used as an input data for the analysis– Implemented data analysis algorithm in Big Data analysis Environment with Spark

– Developed cleansleep.me , a tool which performs natural language processing on TripAdvisor reviews toinform the user of a hotel’s hygiene status by building up Naïve Bayes classifier using Numpy, Scipy,NLTK, scikitlearn , pandas , BeautifulSoup – Deployed an interactive front-end using Flask, MySQL , Bootstrap, and AWS

– Analyzed terabytes of scientific data collected by ATLAS for SUSY and Higgs Boson research– Managed distributed data system used for the multi-dimensional cluster database for High Energy Physics– Developed Deep Learning algorithm in Python to classify deep physical features in particle decay processes

– Generated terabytes of Monte Carlo simulated data which enabled investigation of sources of error on the measurement by quantifying and propagating the errors associated with physics uncertainties in these simulations– Contributed to legacy code refactoring (C++ and ROOT) to achieve a 5x speedup in code execution time while reducing need for manual intervention and job resubmission to batch system

– Held main responsibility of performing final statistical analysis to determine significance of signal over background noise (maximum likelihood fit) in the Higgs boson (H→ WW*) study– Engineered statistical infrastructure to properly account for correlations between parameters in the different regions, resulting in a 5% improvement in analysis sensitivity– Performed analysis which led to discovery level significance in this channel of Higgs decay for the first time in the ATLAS experiment– Published peer reviewed papers, including one in Physical Review D, with more than 1300 citations in total– Developed analysis algorithm to check quality of hardware before installation into ATLAS subdetector, Tile calorimeter in Python