Varshini Periyasamy

ClickStream Customer Intent Prediction

• Developed a customer intent prediction model using large-scale clickstream data, leveraging sequence modeling, and NLP techniques.
• Implemented graph-based community detection (Louvain algorithm) and embedding techniques to automate intent discovery and mapping.
• Built and classified key customer intents, including purchase, product comparison, and order management.
• Achieved 95% classification accuracy, significantly improving intent prediction reliability.
• Generated actionable intent insights that enhanced customer targeting, personalization, and engagement strategies.

MDS Agent – Job Failure Analysis & Automated RCA

• Designed and developed an MDS Agent for automated job failure analysis by integrating Splunk connectors for real-time log ingestion and monitoring, along with a GitHub MCP server to dynamically retrieve relevant code, configurations, and recent commits.
• Leveraged LLM-based reasoning and context-aware prompting pipelines to correlate logs, metrics, job metadata, and code changes, enabling accurate automated root cause analysis (RCA).
• Delivered actionable failure summaries, probable root causes, and remediation recommendations, significantly reducing mean time to resolution (MTTR), and enabling self-serve troubleshooting for engineering and operations teams, thereby improving overall system reliability and operational efficiency.

Cash Flow Forecasting Using Time Series Models

• Developed a time series forecasting solution using ARIMA, VARMA, and VARIMAX models to predict future cash flows for corporate clients.
• Enhanced financial planning and liquidity management by accurately forecasting short-term cash inflows and outflows.
• Automated the model deployment pipeline, allowing real-time cash flow updates for decision-making.
• Tools: Python, Statsmodels, Excel, SQL.

Loan Default Prediction Model

• Developed an end-to-end classification model to predict loan defaults using historical customer data (demographics, credit scores, and loan history).
• Processed large datasets, performed feature engineering, and built the model using logistic regression, random forest, and XGBoost algorithms.
• Deployed the model to production, improving loan approval processes, and reducing default rates by 15%.

Temenos Retail & Corporate Banking (TCIB)

• Experience in Digital Banking products like Retail & Corporate (TCIB, TCMB, OLB), with entire T24 coverage, and Mobile App Testing.
• Involved in the functional testing for TCIB Retail & Corporate modules.