ASHISH CHOUDHARY

Build Survival Model analyzing churns of enterprise level data using PySpark on different cohorts for Marketing team to estimate CLTV. Working on a Risk project identifying customers likely to default in the next 90 days and building Credit Scorecard to expand the product line to customers increasing revenue and mitigate losses. Developing pipelines in Python for analyzing data distribution, metrics, stability index for automation in insights and to reduce project time by more than 25%. Validating the models running in production by checking on stability metrics, drifts and recalibrating the predictive models if needed. Managing a team of 2-3 Consultants for deliverables and mentoring in data science project, statistical modelling and ad hoc analysis.

Worked with cross functional teams to make AI Product combining ML and NLP to predict fraudulent billings for Legal Insurance clients reducing expenses and improving model metrics by more than 10%. Identified key metric to assess Law Firm behavior and developed a methodology around metric for internal stakeholders for better selection and avoiding least risky invoices. Word Embeddings, Syntactic Parsing NLP techniques were used to extract information from unstructured key feature. Contributed along with SME and stakeholders to avoid bias, make product more robust and identify duplicate claims and quantify the cost savings for a very big Insurance firm. Developed Predictive Models using XGBoost and worked with Engineering team to deploy models in production and responsible for model recalibrating using feedback mechanism.