ANKIT GUPTA

Promotion Optimizer: Led development of a promotional optimization system that enables Category Managers to create, and optimize promotional calendar for promotion planning (slotting). Tool provides near real-time sales forecasts to optimize promotional planning and supplier negotiations.

Key Contributions:-

• Implemented end-to-end ML pipeline using XGBoost models for each retail category
• Coordinated three-stage model review process and successful deployment using GCP Vertex AI

Key Achievements:

• Delivered $16M validated gross profit improvement in H2 2023, against identified opportunity of $29M.
• Continuous improvement in NPS score, adoption rates and user engagement proving usability of product.
• Technical Skills: ML Pipeline Development, GCP Vertex AI, Docker, ML modelling (Final Approved model XGBoost), Git Version control for development/deployment.

Enterprise Analytics Platform Transformation: Spearheaded the transformation of key retail analytics platforms, implementing GenAI and standard analytics practices to improve efficiency, and deliver tangible business results:

• Automated Promotion Governance investigations by developing and deploying an GenAI Agent with GCP data warehouse integration, reducing monthly ad-hoc investigations from 25 to 5.
• Implemented CI/CD framework with automated testing, reducing existing deployment time by 60%.
• Streamlined bi-annual refresh cycles through comprehensive run-sheets and automated quality frameworks

• Automated service-request process in customer support system for Asian client using NLP and deep learning (Bi-Directional LSTM) which saved 2 FTEs resources
• Implemented MLOps practices using mlflow, docker, and tensorflow-serving for cost-effective deployment and monitoring
• Developed forecasting solution using LSTM Neural Network for market linked insurance product reserving
• Deployed real time prediction-dashboard through python-dash (flask api) and mysql DB using NinjaTrader indicator data on client's server
• Transformed existing non-life insurance pricing tool into a Python-based solution, creating a more cost-effective alternative to Radar/Emblem while implementing end-to-end machine learning for pricing
• Delivered comprehensive machine learning training program to US client covering statistics, data preparation, supervised ML using Python, and model development/deployment on Azure DataBricks

• Company Overview: Pune
• Developed and deployed multiple propensity models: Channel propensity model (XGBoost) on Azure DataBricks to rank digital channels for existing customers, reducing cost per lead by 30% for personal loan campaigns
• Deployed XGBoost for FD to optimize email targeting (Lift: CTR 1% to 8%)
• Deployed SVM for small personal loans to identify hot leads, improving email campaign CTR
• Designed and implemented data pipeline to automate batch processing of leads from Google Drive to Azure data warehouse using Python API, reducing processing time by 48 hours
• Implemented daily campaign programs for FD and Salaried Loan by analyzing customer intent through Google Analytics clickstream data (Lift: CTR 1% to 13%)
• Pune

• Implemented analytics solution for procurement optimization: Applied Association Rule Mining in R to support category managers in product purchase decisions
• Developed ARIMA forecasting models to improve budget estimation for purchase orders
• Spearheaded development of cost-saving applications, including an interactive R Shiny application leveraging NLP to extract, summarize, and visualize critical information from PDF technical manuals

Insurance Analytics:
o Developed "Dream Retirement" pension product for Norwegian BFSI client through secondary market research, Monte-Carlo simulation, and K-means clustering in Microsoft Azure ML and R
o Created disability insurance RTW prediction model as a new product offering to reduce long-term claim amounts and optimize risk assessment

• 
  Customer Complaint analytics: Delivered (Ticket Analysis and Prediction Engine) solution for multiple clients: -
  Implemented linear regression models to predict complaint resolution time
  Developed time series forecasting to predict complaint volumes
  Applied NLP and text mining to analyze root causes
  Added K-means clustering for education sector complaint segmentation

• Company Overview: Rajasthan
• Data analysis of online system parameters pertaining to production and execution of prescriptive recommendations to ensure availability of system and improvement of production process
• (Implementation of Machine learning: Decision Tree to predict failure of oil-well)
• Accountable for planning and executing electrical operation activities that include forecasting of oil production in R studio, troubleshooting of issues and preparation of periodic reports
• Rajasthan