S Karthikeyan

Data Scientist 2 @ Swiggy

About

As a Data Scientist at Tiger Analytics, I specialize in building production-grade solutions using Generative AI, LLM-based intent classification, and RAG architectures. My recent work includes enhancing chatbot performance with GPT-4 Turbo and FAISS-powered fallback recovery, improving user experience through better intent resolution, and optimizing customer insights using unsupervised techniques like HDBSCAN and UMAP. Previously, I contributed to Shell's data-driven transformation as an Associate Data Scientist by developing clustering models and semi-supervised BERT pipelines that reduced manual effort and improved accuracy across large-scale datasets. My early foundation was shaped at LTI Mindtree and Kanini, where I deployed real-time fraud detection systems and built scalable recommender engines. Armed with a B.E. from the Government College of Technology, Coimbatore, and advanced certifications from Imarticus, UCLA, and Scaler, I combine deep technical knowledge with a product-focused mindset to deliver business impact through AI.

Country

India

City

Chennai

Industry

Computer Software

Skill

Small Language Models (SLMs), Retrieval-Augmented Generation (RAG), qlora, BERT (Language Model), Fraud Detectiion Modelling, Recommender Systems, Transformers, Cross Encoderss, API Integeration, Cloud Computing, Agentic AI, Amazon Web Services (AWS), Cluster Analysis, Continuous Integration and Continuous Delivery (CI/CD), Generative AI, Responsiveness, Data Ingestion, Process Mining, Celonis, Boosting

Experience

Data Scientist 2

Swiggy

2026-2 - Present · 8 mos

Bengaluru, Karnataka, India

Data Scientist

Tiger Analytics

2024-8 - 2026-2 · 1 yr 7 mos

Chennai, Tamil Nadu, India

• Designed and deployed a robust enhancement pipeline for chatbot performance and fallback recovery. • Utilized HDBSCAN clustering with PCA/UMAP for intent overlap resolution, boosting accuracy by 8%. • Architected a RAG-based fallback system on Amazon Lex, improving ambiguous query handling significantly. • Automated bot testing via AWS Lambda, ensuring consistent quality across versions.

Associate data scientist

Shell

2023-7 - 2024-6 · 1 yr

Chennai, Tamil Nadu, India

• Enhanced operational efficiency by clustering users based on invoice features using K Prototype and DBSCAN algorithms. • Achieved a 30% reduction in invoice processing time and a 15% improvement in overall efficiency. • Developed a semi-supervised learning pipeline, automatically labeling over 150K support tickets, reducing manual workload by 80%. • Contributed to Shell's innovative approach in data science to optimize support operations in Chennai, Tamil Nadu.

Associate Data scientist

LTI - Larsen & Toubro Infotech

2022-5 - 2023-7 · 1 yr 3 mos

Chennai, Tamil Nadu, India

• Developed and deployed a robust fraud detection system utilizing Isolation Forest and Local Outlier Factor techniques. • Implemented Docker for scalable and consistent deployment, enabling real-time monitoring of transactions. • Created a personalized product recommendation system that increased customer engagement by 25% and improved sales conversion rates by 18%. • Contributed to LTI's mission of leveraging data science to enhance business operations and customer experiences.

Junior Data Scientist

KANINI

2021-10 - 2022-5 · 8 mos

Chennai, Tamil Nadu, India

• Developed and deployed end-to-end machine learning pipelines for attrition prediction, reducing customer churn by 22%. • Designed scalable data processing workflows for real-time feature ingestion and model retraining. • Implemented the full MLOps lifecycle using Docker and GitHub Actions to ensure automated deployment. • Created interactive Streamlit applications for stakeholders to visualize predictions and take proactive actions.