Manisekhar Kothuri

Career Progression: Intern (Jan 2020 – Jul 2020) → Data Analyst (Aug 2020 – Jul 2021) → Senior Data Analyst (Aug 2021 – Jul 2022) → Data Scientist (Aug 2022 – Jul 2024) → Senior Data Scientist (Aug 2024 – Present)

Generative AI Chatbot – US Insurance Client

• Developed and launched Generative AI chatbot for fee schedule and notices leveraging semantic routing and multi-agent frameworks through Langraph and OpenAI LLMs.
• Enhanced reliability of chatbot through LLM-driven validation layers and dynamic routing strategies for query classification.
• Built automated document ingestion pipeline, ensuring continuous updates to knowledge base for context-aware RAG workflows.
• Achieved 87% accuracy in user validation and 89% in LLM-based evaluation, reducing average response time from 14 seconds to 10 seconds.

Demand Forecasting

• Developed demand forecasting model for beauty product sales using AWS SageMaker and custom XGBoost objective functions.
• Achieved 12% WMAPE, exceeding baseline prediction accuracy by 60% through advanced feature engineering.
• Mentored and managed two junior data scientists to enhance team performance and deliverables.

Aftersales Churn Prediction

• Engineered multi-market customer churn prediction system with LightGBM, achieving AUC of 0.8 by integrating behavioral, cost, and geospatial data.
• Incorporated Google Maps API and GeoPandas for enhanced predictive accuracy through location intelligence.
• Conducted extensive exploratory data analysis and hypothesis testing using PySpark for large-scale data processing.
• Developed robust data cleaning and transformation pipelines to ensure high data integrity.
• Deployed end-to-end weekly inference pipeline in Azure Data Factory for automated churn risk updates.

Lead Scoring

• Developed real-time lead scoring engine using XGBoost, achieving AUC of 0.9 for vehicle purchase prediction.
• Integrated datasets from Salesforce and Google Analytics to create enriched feature sets.
• Automated deployment via Azure Data Factory, providing real-time scoring to sales teams.
• Boosted online sold rate from 1–2% to 5% through targeted identification of high-conversion leads.

Model Monitoring Dashboard

• Created comprehensive model monitoring dashboard featuring tabs for results, prediction tracking, and data drift detection.
• Facilitated interactive scenario testing through What-If Tool to enhance client trust in ML outputs.
• Integrated continuous performance tracking and drift analytics into production workflows for proactive model retraining.