Browse categories
Explore
Fiverr Pro
English
$
USD
I will build custom ai agent pipelines and multi agent automation workflows
Saad Abdullah
About this gig
Tired of one-shot LLM calls that don't actually automate anything? I build multi-agent systems where AI agents plan, execute, communicate, and self-correct end to end.
I specialize in agentic AI architecture. My work includes a 5-agent automated video generation pipeline (Story Audio Video Edit Undo), built with an MCP tool-abstraction layer, LangGraph-style orchestration, Pydantic state contracts between agents, a WebSocket-live FastAPI backend, and a natural language edit + version revert agent 46/46 tests passing.
What I can build for you:
- Multi-agent pipelines with Supervisor/Worker or sequential architectures
- LangGraph workflows with custom node logic, edges, and conditional routing
- MCP (Model Context Protocol) tool servers and agent-tool integration
- Agents that call external APIs, process files, query databases, or control other tools
- State management with versioning and rollback between agent phases
- Natural language intent classification feeding into automated action planners
- FastAPI backends exposing your agent pipeline as a REST/WebSocket API
- LLM-provider-agnostic systems (OpenAI, Gemini, Claude, Ollama swap without rewrites)
Get to know Saad Abdullah
Saad Abdullah
- FromPakistan
- Member sinceFeb 2024
- Last delivery2 years
Languages
English
I specialize in Agentic AI, GenAI, and MLOps. Built production-grade multi-agent systems with LangGraph and MCP, deployed Stable Diffusion inference on GPU cloud (RunPod), and shipped end-to-end ML pipelines with Airflow, MLflow, and Docker. Seeking an AI/ML Engineer role to own LLM-powered systems from architecture to deployment. I also do 2D/3D Unity game development and have expertise in C#
FAQ
Do I need to provide an OpenAI / Gemini API key?
You'll need a key for whichever LLM provider you want to use. I build all systems to be provider-agnostic — if you switch providers later, it's a one-line config change, not a rewrite. I can also build with Ollama for fully local/offline setups.
What information do I need to give you to get started?
A clear description of what you want the agent(s) to do: the input, the desired output, any tools or APIs the agent should use, and any constraints (offline-only, specific LLM, etc.). The more specific, the better.
Can you integrate with my existing codebase?
Yes, message me first with a brief description of your stack and I'll confirm compatibility before you order.
Do you use LangGraph specifically?
I implement LangGraph-style orchestration and can use LangGraph directly or build equivalent graph logic in plain Python (no extra dependency). Your call.
Will I be able to understand and extend the code after delivery?
Yes. I write production-quality code with docstrings, a README, and where applicable, tests. I won't hand you a monolithic script no one can maintain.
What does "MCP tool layer" mean and do I need it?
MCP (Model Context Protocol) is a way to expose tools to LLM agents in a structured, reusable way. It's useful for larger systems where multiple agents share the same tools. For simpler single-agent setups it's optional — I'll recommend the right architecture for your scope.
