Agentic systems
01LangGraph agents that plan, call tools and hand off — with guardrails, retries and human review where it matters.
- LangGraph orchestration
- Tool calling
- Evaluation harness
Muhammad Mubeen
Full-Stack Agentic AI Engineer. Full-stack web builds, RAG chatbots, custom AI automation and the production systems around them 11 shipped and counting.
1// route the request, then let the graph decide2const graph = new StateGraph(AgentState)3 .addNode("retrieve", retrieveContext)4 .addNode("answer", callModel)5 .addConditionalEdges("retrieve", needsTool)6 .compile();78const run = await graph.invoke({ question });Production systems with a problem, a stack and an outcome. Each one links out to the live build.
Chosen for boring reliability, not novelty. Every one of these is running in something I shipped.
Frontend
Backend
AI
Infra
Fixed-scope builds, retainers, or an audit of an AI feature that isn't performing.
LangGraph agents that plan, call tools and hand off — with guardrails, retries and human review where it matters.
Retrieval grounded in your data: chunking, embeddings, pgvector search and answers that decline instead of inventing.
Low-latency speech agents for support and intake — streaming STT, barge-in, and a transcript your team can audit.
The whole product around the model: Next.js or NestJS, Postgres, auth, billing, background jobs, deployed and monitored.
Let’s talk
Tell me the problem and I’ll tell you whether an agent is the right answer. Or ask my digital twin in the corner — it knows my work.