GAUTAM KUMAR

AI Engineer and Senior Data Scientist with over 12 years of experience in developing production-grade GenAI and full-stack data science solutions. Expertise in advanced LLM techniques, statistical modeling, and data visualization, driving impactful business decisions. Proven ability to convert research into scalable products while optimizing performance and cost. Recognized for high productivity and efficiency in delivering actionable insights through machine learning and predictive analytics.

• Top 5% in multiple hackathons: WNS Analytics Wizard (2019), Club Mahindra Data Olympics, and LTFS Data Science FinHack
• GitHub contributor: repositories demonstrating RAG pipelines, LoRA fine-tuning examples, and production MLOps patterns

• Date of Birth: 12/18/90
• Nationality: Indian

• English
• Hindi

Cisco GenAI-powered organizational hierarchy validation (agentic AI) built an agentic pipeline using LangChain, Langraph, and search APIs to autonomously validate org structures. Components include web scraping, evidence retrieval, confidence scoring, human-in-the-loop review, and using AWS Bedrock 

Sharecare Customer engagement analytics, developed a scalable end-to-end call analytics platform integrating Whisper transcription, speaker diarization, sentiment and empathy detection, and LLM-based summarization Applied LoRA fine-tuning on Llama 3.1 models, implemented monitoring and feedback loops, and deployed using Docker on AWS with CI/CD pipelines Advanced RAG Summarization & QA System, Engineered hybrid retrieval (semantic + sparse), vectorDB integration, chunking strategy, and LoRA-based fine-tuning of reader models to improve domain QA and reduce hallucinations.

STC B2B — churn and issue prediction, led churn and issue prediction modeling for B2B accounts using SAS Viya, Python, and Teradata; delivered dashboards and operational workflows enabling proactive retention strategies,

SeeR Analytics Platform- delivered multiple production use cases (Order-to-Activate, Invalid Truck Roll, Ticket Aging) on an AWS-hosted analytics platform; implemented orchestration scripts, Lambda functions, and dashboards for operational monitoring

• Architected hybrid RAG systems integrating FAISS/Pinecone with sparse retrievers and chunking strategies; applied LoRA/PEFT fine-tuning to reader models, improving answer relevance and reducing hallucinations.
• Built agentic AI workflows for automated org-hierarchy validation and QA automation with human-in-the-loop review, confidence scoring and audit trails.
• Led speech analytics: Whisper transcription, speaker diarization, sentiment & empathy detection, and LLM summarization for post-facto and real-time insights.
• Established MLOps practices: MLflow model registry, CI/CD, containerized deployment (Docker/K8s), monitoring, drift detection, latency & cost optimisation.
• Delivered end-to-end analytics solutions (ETL → modelling → dashboarding) that improved operational KPIs and enabled proactive interventions (churn reduction, repeat-call mitigation).

• Cricket and Coffee

• Tech Mahindra ACE Award - 2024