Tejan Arora

1.First Call Resolution(FCR) model - Developed an FCR prediction model using XGBoost to enhance client experience and contact center efficiency. Leveraged call transcripts, metadata, and sentiment data to predict whether an agent could resolve a query on the first call. Implementation led to improved customer satisfaction rates and a higher FCR score, driving operational efficiency.

2. Licensing and Compensation Project- Designed an NLP-driven solution to reduce call handle time by transferring calls to IVR and chatbots. Used speech-to-text conversion and topic modeling to categorize calls, extract context, and identify transfer opportunities. Additionally, built a machine learning model to predict call frequency on a monthly basis, enabling better workforce planning. This optimization reduced 2–3 US FTEs, leading to a cost savings of $500K while improving efficiency.

3. Synthetic Data Generation - Developed this utility to enhance data availability and privacy within the organization. For tabular data, we leveraged the CTGAN Deep Learning library to generate high-quality synthetic datasets that maintain statistical integrity. For textual data, I utilized the Mistral7B parameter LLM Gen-AI model to create realistic synthetic text while preserving contextual relevance.

=> This solution effectively solved a business problem, and multiple teams across the organization now use it for data-driven workflows.

4. Risk Model for Underwriting Process - Designed and implemented an end-to-end MLOps pipeline for an underwriting ML model in the insurance domain, leveraging AWS SageMaker to automate data ingestion, model training, validation, deployment, and monitoring. Built robust pipelines using SageMaker Pipelines for preprocessing and model retraining, integrated CI/CD workflows with GitHub Actions and CodePipeline for seamless deployment, and utilized SageMaker Model Monitor and CloudWatch for real-time performance and drift monitoring. The solution enabled scalable, real-time risk assessment during underwriting, significantly reducing manual effort, improving model reliability, and ensuring regulatory compliance through full traceability and version control.

5. ETF Prediction Model - Developed a predictive model using AWS SageMaker AutoML to forecast whether an ETF price would change by more than 4% in the next 4 weeks. After multiple iterations, achieved ~80% accuracy with strong precision and recall, enabling better investment decision-making.

• Optimized the client's collection strategy, especially during the prevailing COVID-19 pandemic
• Identified High-Risk Account in the complete portfolio using Advanced ML techniques like Random Forest
• Optimized the process for credit Line Assignment for the Client
• Early Account management for the high-risk accounts
• Risk Analytics

• Analyzing day-to-day leads and creating reports, dashboards, and funnel charts for management
• Segmenting customers and providing them with customized offers using the K-Means Clustering technique
• Forecasting monthly & quarterly sales using Time Series analytics