Satyam Singh

SDE

Sasefied • Temps partiel

Dec 2025 - Present • 5 mos

Designed and developed a full-stack multi-agent AI platform (Sasefied) with a router-based backend that intelligently coordinates multiple AI agents to solve complex tasks. Implemented specialized domain agents (client interaction, compliance validation, optimization, process execution, and supply-chain reasoning) and defined structured workflows to ensure reliable and traceable agent execution. Built a synthesizer layer that aggregates outputs from multiple agents into a unified, actionable response, focusing on modular design, scalability, and production readiness.

AI Engineer Intern

FoodNest Technology • Temps plein

May 2025 - Jul 2025 • 2 mos

Designed and implemented a hierarchical multi-level agent architecture using LangChain + LangGraph, featuring orchestrator, planner, executor, validator, and tool-specialist agents that collaborate via stateful graphs to handle complex, multi-step reasoning and task decomposition. Built an end-to-end agentic data processing pipeline with FastAPI, where autonomous agents dynamically route, parse, chunk, validate, and index data across multiple file formats (PPT, DOCX, PDF, TXT, JSON, Excel, etc.), ensuring fault tolerance and scalable ingestion. Enabled multi-source intelligent ingestion through agent-driven connectors, allowing agents to autonomously fetch and normalize data from Google Docs, Google Sheets, static web pages, and arbitrary URLs, with context-aware preprocessing and metadata enrichment. Integrated role-aware RAG with policy-enforcing agents, implementing multi-level access control where retrieval, ranking, and response-generation agents adapt outputs based on user roles, permissions, and organizational hierarchy. Extended the agentic response layer with multimodal output agents, capable of deciding when to generate structured visual explanations (SVG charts/graphs) versus natural language responses, improving interpretability and decision support.