Debashish Panda

About Debashish Panda

Automation Architect @Sundew Solutions | Agentic AI & LLM Specialist | Founder @…

Experience

1. Automation Architect

   Sundew

   May 2024 — Present · Kolkata, IN

   Architected and deployed multi-agent AI systems using LangChain, LangGraph, and CrewAI, enabling task decomposition, coordination, and execution across enterprise procurement workflows.Built scalable RAG (Retrieval-Augmented Generation) pipelines leveraging FAISS, Pinecone, and Snowflake, integrated with structured and semi-structured procurement data sources.Fine-tuned and prompt-engineered LLMs (OpenAI, Claude, open-source) to serve use cases like procurement copilots, automated research agents, and document summarization tools.Developed orchestration logic using LangGraph and CrewAI to coordinate agent roles, memory management, and response routing in multi-agent settings.Integrated LLM agents with enterprise APIs (REST/GraphQL), ServiceNow, and databases for seamless two-way interaction with business systems.Containerized services using Docker, implemented CI/CD pipelines via GitHub Actions, and ensured secure deployment across AWS cloud infrastructure.Managed the full MLOps lifecycle including model versioning, drift monitoring, feedback loops, and rollback mechanisms using tools like MLflow.Applied data privacy techniques such as redaction, anonymization, and access-controlled data flow to ensure compliance with procurement data regulations.Designed and executed experiments for model benchmarking, latency and cost optimization, and A/B testing, regularly presenting findings to product and executive stakeholders.Collaborated cross-functionally with product, design, and customer-facing teams to translate vague requirements into stable, production-ready agent pipelines.

Education

• Government College of Engineering, Keonjhar

  Bachelor of Technology - BTech, electrical engineering

• U.N college, soro

  intermediate, Stem