k
kapilyaduwanshi

Kapil Yadav

@kapilyaduwanshi

AI Agents Systems Engineer

India
English, Hindi
About me
AI Agent Engineer with 5+ years of software engineering experience building production-ready AI agents, multi-agent systems, and RAG applications using LangGraph, LangChain, Python, and FastAPI. I specialize in autonomous workflows, tool calling, MCP integrations, vector databases (Pinecone, Chroma, FAISS), prompt engineering, and scalable AI architectures. I build reliable, enterprise-grade AI solutions that automate workflows, integrate with APIs, and deliver measurable business value. ... Read more

Skills

k
kapilyaduwanshi
Kapil Yadav
Offline • 
Average response time: 1 hour

See my services

Custom GPT Apps
I will build custom ai agents and multi agent systems
From$40 / projectMore details

Portfolio

Work experience

Fiverr

Gen AI Developer

Fiverr • Full-time

Jul 2026 - Present • 0 mos

• Architected and deployed autonomous multi-agent AI systems using LangGraph and LangChain, implementing task decomposition, tool-calling, and decision-making workflows that reduced manual sourcing efforts. • Built production RAG pipelines using Pinecone and Chroma vector databases, enabling semantic search and context-aware LLM responses. • Designed human-in-the-loop checkpoints within agent workflows to ensure safe, auditable autonomous execution for business-critical tasks. •Integrated and developed MCP (Model Context Protocol) for standardized tool and data-source connectivity across multi-agent workflows, simplifying agent-to-tool integration. • Built and deployed scalable AI microservices via FastAPI, exposing agent orchestration, chat, and summarization capabilities to downstream systems. • Deployed and orchestrated containerized AI agent services using Docker and Kubernetes, improving scalability and deployment reliability. • Implemented token-level streaming via SSE(Server Sent Events), async architectures, improving perceived response latency. • Applied structured prompt engineering (few-shot, tool-calling, structured output) to reduce hallucination rates and improve output reliability in production. • Optimized embedding/similarity search using Pinecone, improving retrieval efficiency across high-dimensional datasets.