Santosh Narayanan

• • Smart Developer Assistant (SDA): Architected and deployed multi-agent AI system for automated code intelligence, refactoring, and DevOps artifact generation using LangChain, LangGraph, FastAPI, Qdrant, React TypeScript. Implemented ingestion-to-deployment pipeline processing 50+ repositories with 90%+ accuracy in generating Docker, Kubernetes, and CI/CD templates. Graph + Designed orchestration architecture, caching. (GitHub

• • Designed and implemented Agentic AI workflows using LangChain and LangGraph to automate complex financial data-mapping processes across 10–20 client-specific workflows, reducing manual effort significantly.

• • Built Retrieval-Augmented Generation (RAG) pipelines with LangChain and Pinecone to enable LLM driven semantic search and rule-based data transformation for banking templates.

• • Developed AI-assisted data ingestion pipelines capable of dynamically adapting to client-specific schemas, rules, and column structures, persisting changes automatically in the database.

• • Implemented an Agentic AI-driven DevOps automation system that analyzed frontend/backend repositories (TypeScript, JavaScript, Angular) and auto-generated Docker Compose configurations, reducing setup time from days to minutes.

• • Refactored backend services to FastAPI + SQLAlchemy, improving memory efficiency by 45%, and containerized full-stack services for production-ready CI/CD deployment.

• • Designed deterministic agentic workflows with structured outputs, rule-based validation, and schema enforcement to evaluate correctness, prevent hallucinations, and ensure reliable execution across multi-step RAG and tool-using pipelines.

• • Implemented multimodal agent pipelines integrating OpenAI Whisper for speech-to-text ingestion, enabling voice-driven inputs to be processed through LangChain-based RAG and multi-agent reasoning workflows.

• • Designed and implemented asynchronous processing pipelines using FastAPI with RabbitMQ and Celery, improving system scalability and reducing processing bottlenecks while increasing workflow throughput by ~35%.

• • Implemented proactive monitoring and quality controls using Grafana dashboards and automated pytest suites, reducing production issues and improving system stability and mean-time-to-resolution.

• • Built ReactJS components with AWS Cognito & S3, increasing engagement 15%.

• • Optimized GraphQL with DataLoader, improving performance by 35%.

• • Strengthened backend security and reliability by implementing JWT-based authentication, FastAPI middleware, and API hardening patterns, improving platform robustness and reducing operational issues in production environments.