Gowri Saravanan

AI Engineer with almost 3 years of experience building enterprise Generative AI applications and AI platforms at TCS Research & Innovation (CTO Unit). Experienced in developing end-to-end AI solutions spanning LLM applications, advanced RAG and agentic workflows, FastAPI microservices, React interfaces, model serving, evaluation, and Azure-based deployments. Hands-on with LangChain, LangGraph, MCP, FAISS, hybrid retrieval, reranking, LlamaFactory, and vLLM, with experience translating AI research into scalable enterprise solutions.

Systems Engineer (AI Platform & Generative AI) - Tata Consultancy Services – Research & Innovation (CTO Unit)

(2024-01)

• Architected backend services for an enterprise AI/LLM lifecycle platform supporting dataset onboarding, fine-tuning, evaluation, inference deployment, checkpoint management, and training monitoring for internal data scientists.
• Built an orchestration layer over LlamaFactory to standardize enterprise fine-tuning workflows and expose model lifecycle operations through scalable APIs.
• Engineered asynchronous AI workflow and scheduling services for long-running training, evaluation, deployment, and batch execution, with dependency validation, status tracking, cancellation, and resume capabilities.
• Deployed and optimized vLLM-based inference on Azure GPU infrastructure for LoRA-enabled LLM serving, supporting scalable enterprise inference workloads.
• Developed a centralized model evaluation framework exposing BLEU, ROUGE, and Perplexity metrics, alongside AI-assisted hyperparameter recommendations and automated training summaries.
• Built a Graph-based Policy RAG pipeline using FAISS, query rewriting, confidence-threshold gating, reranking, and graph-aware retrieval to improve retrieval precision for enterprise policy knowledge.
• Designed a metadata-aware semantic recommendation engine using vector retrieval and ChromaDB to recommend reusable enterprise workflow activities from historical knowledge.
• Designed and maintained ~7 backend microservices exposing 100+ REST APIs, with asynchronous execution, Docker-based deployments, Keycloak authentication, and RBAC.
• Implemented backend services and React integration for a Graph-based Policy RAG platform, contributing to its transition from a research PoC to an internal production capability.
• Integrated TransformerLens-based LLM explainability into the enterprise platform through backend APIs, model preparation workflows, persistent trace storage, and an AI-assisted interpretation interface.
• Collaborated with Customer Success and Account teams to translate business requirements into enterprise AI workflows and configure AI agent solutions.

B.E. - Instrumentation & Control Engineering - Sri Sairam Engineering College, Chennai (2019 - 2023)