Bhavesh Sidhwani

Software Engineer at CareerUS Solutions

Led the development of an AI-powered slot-filling recruitment chatbot and a multi-agent LangChain system for query orchestration. Implemented context management to preserve conversation history and job metadata across sessions. Integrated MCP for job posting retrieval and automated creation with a human-in-the-loop workflow, resulting in faster responses and improved posting accuracy. Tech stack included Python, LangChain, TypeScript, React, and Node.js.

Software Engineer at Cielara AI

Built a RAG pipeline powering seven AI agents for infrastructure security, management and compliance. Developed an infra-scanner Go service that converts observability data into a knowledge graph persisted on GraphDB, and integrated a vector DB to enable subgraph retrieval, reducing token usage by 75%. Designed a continuous evaluation pipeline (RAGAS) to monitor agent performance and prevent model drift. Engineered a dual-LLM critique feedback agent to boost relevancy evaluation.

Full Stack Developer at Saayam for All

Developed responsive web pages with React and Redux, implemented scalable microservice APIs with Spring Boot, and integrated AWS Cognito for secure authentication. Improved UX across devices and browsers with a focus on accessibility and performance.

Software Developer Intern at MAD Engineers

Designed an Inventory Management System (Java, Android, OAuth2, Firebase) that boosted operational efficiency and reduced stockouts. Migrated legacy data from Google Drive via OAuth2 and Drive API, and secured data with Firebase Authentication and Realtime Database. Enhanced user experience with lazy loading and optimized memory usage.

Software Engineer at Cielara AI

Built AI-native cloud reliability platform with Next.js frontend, Go backend, and Python AI microservices on GCP. Developed multi-cloud infra scanner in Go to construct digital twin graphs with 50K+ nodes persisted on FalkorDB. Architected modular agent framework, reducing new agent development time from 2 weeks to 4 days. Implemented REST APIs for infra operations and WebSocket APIs for agent interactions with sub-200ms latency. Designed MongoDB session management with hierarchical summarization, enabling 35% longer conversations. Reduced RAG token usage by 75% through MilvusDB vector store with targeted sub-graph retrieval.