Data Scientist Associate Principal
Accenture
Total years of experience :6 years, 11 Months
• Led the implementation of proof-of-concepts (POCs) and collaborated closely with business units to deliver the developed
product to five clients. Managed a team of 3 junior analysts, guided them in implementation of features and corresponded with
cross-functional teams to ensure seamless product integration within client environments.
• Worked from end to end in development of product by coordinating with multiple business units to gather requirements,
conceptualizing the flow for anomaly detection in time series data using graph networks, and successfully implementing the
solution utilising Networkx for building graphs which identified fraud entities 16% better than structured model.
• Spearheaded an innovative solution approach using Generative AI to generate recommendations to mitigate anomalies from time
series data. This cutting-edge approach allowed for on-the-fly inferences, resulting in a substantial 75% decrease in storage space
requirements.
Working in Product Development Team, developed 2 Anomaly Detection modules, conceptualized Root cause Identification and
Impact Analysis modules and deployed the production ready codes as micro-services with the help of back-end team by
adhering to the best practices in coding.
• Owned 6 models end-to-end for predicting fraud, subrogation, litigation, and severity for a leading US insurance company, utilizing
machine learning techniques such as clustering, regression and classification using random forest, GBM, and XGBoost
models.
• Created materials and coached 120 fresher’s in “supervised algorithm” concepts such as Logistic Regression, Decision Tree
models, concept of Bagging and Boosting, ensemble techniques and Basics of Neural Networks.
• Collaborated with an insurance client to enhance their Fraud Detection model. Utilized Link Analysis techniques to extract
graph features and incorporated them into a supervised model using XGBoost which increased the model's recall from 65%
to 81%, resulting in a five-fold reduction in claims investigation time and substantial cost savings.
• Created streamlined solution for exploratory data analysis (EDA), feature selection, and dimension reduction tailored
specifically for the Finance business unit using Pandas, Numpy and sklearn. This solution led to a remarkable 50% acceleration in
project delivery, enabling faster insights and decision-making.