Senior Data Scientist - GenAI & RAG

Kranthi Kumar Adepu

Senior Data Scientist

Hyderabad, India ******************@*****.*** +91-810******* LinkedIn Professional Summary

Results-driven Senior Data Scientist with 9+ years of experience designing and deploying scalable AI/ML solutions across Marketing, HR, and Banking, Retail domains. Proven expertise in end-to-end project lifecycle management from business problem framing to model deployment on Azure cloud. Specialized in ML, DL, NLP, Generative AI, LLM evaluation, RAG, AI Agents, Survival Analysis, and handling class-imbalanced problems. Strong collaborator with data engineering teams to build robust data pipelines using Azure Databricks, ADF, and Synapse. Adept at translating complex model insights into actionable business strategies using Power BI. Professional Experience

Commvault Systems — Senior Data Scientist

Hyderabad, India 2021 – Present

Led AI/ML initiatives to solve critical business problems, driving measurable ROI through advanced analytics and GenAI solutions. Deployed end-to-end AI/ML models by building data pipelines and scheduling job triggers daily/weekly.

Key Projects & Achievements:

• Building Multi-Agent Architecture for SolarWinds Service Desk

Developed an enterprise Agentic RAG platform using LangGraph, Azure OpenAI, Azure AI Search, and FastAPI for Slack-based support automation routing user queries through an orchestrator agent to specialized department, retrieval, re-ranking, and response agents.

Implemented multi-agent workflows for ticket lookup (fetching from Azure Blob), semantic retrieval with recursive chunking and embedding model (Ada Small), re-ranking, and GPT-powered response generation via LangGraph tool-calling agents.

Extracted data using SolarWinds APIs, stored in Azure Data Lake, embeddings stored in Azure AI Search with ticket ID and department as metadata; deployed as Docker image on Azure Container Apps.

• RAG Application Building and LLM Chatbot Evaluation for Support Tickets

Designed and implemented data ingestion and retrieval pipelines for Arlie-L1 Support; augmented context to generate ticket responses.

Evaluated Arlie L1-support bot using BLEU, ROUGE, Perplexity, and BERT Score against human benchmarks.

Designed methodology to measure hallucination rates for seen/unseen tickets, improving output reliability by 30%.

Leveraged Azure AI Services as a vector DB for embedding storage and retrieval.

• Marketing: Job Title Classification (BERT + XGBoost)

Built NLP pipeline to classify lead titles into Level/LOB using BERT and Word2Vec embeddings.

Achieved 91% accuracy for Level and 90% for LOB classification.

Deployed model on Azure Databricks; automated data ingestion via ADF pipelines and Azure Synapse.

• HR: Employee Attrition & Time-to-Exit Prediction

Developed 2-tier model (XGBoost + Cox PH) predicting attrition likelihood and time-to-leave.

Achieved 86% classification accuracy and 78% C-index. Used SHAP for transparent feature impact analysis.

• Marketing: Opportunity Closure Time Prediction (Survival Analysis)

Applied Cox Proportional Hazard model to forecast sales cycle duration.

Achieved 86% C-index; segmented models for new vs. existing customers.

Deployed on Databricks with weekly Spark pipeline automation.

• Marketing: Metallic Buyer Prediction (Class Imbalance Handling)

Addressed 99:1 class imbalance using resampling technique (SMOTE) and class-weight parameter; Recall was around 80–85%.

Automated end-to-end workflow via Spark jobs scheduled in ADF. 3LOQ Labs Pvt Ltd — Data Scientist

Hyderabad, India 2019 – 2021

Developed AI-powered recommendation engine for HDFC Bank to boost customer onboarding and engagement. Key Projects & Achievements:

• "Habitual AI" Recommendation & Onboarding Engine

Built modules for transaction categorization, attrition prediction, and customer habit analysis using Random Forest and Cosine Similarity.

Segmented customers into 3 vintage cohorts (0+, 1-2, 2+ months); built separate models for each.

Measured success via MAP (Mean Average Precision) for login rate, transaction count, and variety uplift.

Automated Spark job scheduling via shell scripts; delivered pre/post-campaign analysis showing 15–20% engagement lift.

Presented test vs. control group results to client stakeholders. Diwo – Loven Systems / DataFactZ — Data Scientist

Hyderabad, India 2017 – 2019

Delivered data science solutions for Retail and Banking clients using advanced ML techniques and cloud infrastructure.

Key Projects & Achievements:

• Retail Analytics for LBrands (Cannibalization & Cross-Sell)

Predicted customer revisit propensity using XGBoost and ANN.

Conducted product clustering and analyzed cannibalization/halo effects.

Performed feature engineering and model validation on AWS EC2.

Insights integrated into Diwo's decision intelligence product.

Built time-series forecasting models in Python and R. Education

Bachelor of Technology (B. Tech) in Computer Science & Engineering JNTU College of Engineering, Manthani 2016 67% Key Achievements

• "Best Performance Bronze" Award – Commvault Quarterly TownHall

• "Employee of the Month" – DataFactZ / Diwo – Loven Systems Technical Stack

• Languages: Python, SQL Server, PySpark

• ML/DL: Descriptive Analysis, Predictive Modelling, Decision Tree, Random Forest, ANN, LSTM, XGBoost, LightGBM, BERT, Transformers, Survival Analysis, Recommendation Engine, Association Rule Mining (Apriori), Time Series Models, TensorFlow, PyTorch

• NLP & GenAI: LLM Evaluation (BLEU, ROUGE, BERT Score, Context Relevancy, Answer Correctness), RAG, Hallucination Detection, Tokenization, Embeddings, Word2Vec, AI Agents, LangChain, LangGraph, Agentic AI

• Cloud Platforms & Visualization: Microsoft Azure (Databricks, Data Factory, Synapse, Blob, Data Lake, AI Services, Azure Web Apps, Azure Container Apps) and Power BI

Contact this candidate