AI Engineer · LLM Agents Architect · AI Automation Developer · AI Evaluation Specialist

I build AI systems around LLM agents, RAG, automation, local inference, tool use, evaluation, and self-improving workflows.

I like working close to how AI agents behave: how they understand tasks, use context, choose tools, fail, recover, and improve through feedback.

· GitHub · X · Bluesky

• LLM agents that use tools, APIs, memory, retrieval, and structured workflows
• RAG systems for knowledge bases, internal assistants, and source-grounded answers
• AI automation with n8n, webhooks, APIs, schedules, and approval flows
• Agent identity files such as AGENTS.md, CLAUDE.md, GEMINI.md, llms.txt, and project memory
• Local AI agents using Ollama, LM Studio, llama.cpp, vLLM, SGLang, GGUF, and local APIs
• Evaluation workflows for prompts, agents, retrieval quality, tool calls, and model behavior
• Self-improving agent patterns using feedback, traces, rubrics, memory, and regression tests
• MCP-based tool integrations, coding agents, browser agents, and agentic workflows

I think about AI systems as a loop, not a one-shot prompt.

flowchart TD
    A[Project Identity] --- B[Instructions and Memory]
    B --- C[Context and Knowledge]
    C --- D[Tools and APIs]
    D --- E[Agent Runtime]
    E --- F[Model Layer]
    F --- G[Evaluation and Observability]
    G --- H[Feedback and Improvement]
    H --- B

    C --- C1[RAG]
    C --- C2[Knowledge Base]
    C --- C3[Project Docs]

    D --- D1[MCP]
    D --- D2[Webhooks]
    D --- D3[Local Tools]

    E --- E1[Planning]
    E --- E2[Routing]
    E --- E3[Human Approval]

    F --- F1[Hosted Models]
    F --- F2[Local Models]

    G --- G1[Traces]
    G --- G2[Rubrics]
    G --- G3[Regression Tests]

A strong AI system needs more than a model. It needs identity, instructions, context, tools, memory, evaluation, observability, and a way to improve after each run.

• LLM agent systems
• Local AI agents
• RAG-based assistants
• AI-ready knowledge systems
• Agent instruction files
• AI automation workflows
• Internal AI copilots
• Multi-agent research workflows
• Tool-calling agent infrastructure
• MCP-based tool integrations
• Evaluation and feedback pipelines
• Human-in-the-loop AI workflows
• AI prototypes connected to real tools and data

• Tool-calling agents
• Planner-executor workflows
• Router agents
• Research agents
• Retrieval agents
• Multi-agent coordination
• Agent memory and state
• Agent handoffs
• Human approval flows
• Guardrails and tool validation
• Agent tracing and debugging
• MCP-based tool integration
• Local agent runtime design

• AGENTS.md workflows
• CLAUDE.md project memory
• GEMINI.md instructions
• llms.txt documentation maps
• Cursor rules
• Windsurf rules
• Project memory files
• System prompt files
• Prompt contracts
• Repository-specific agent guidance

• RAG pipelines
• Agentic RAG
• GraphRAG
• Hybrid search
• Query rewriting
• Query expansion
• Reranking
• Context packing
• Context compression
• Source-grounded answers
• Knowledge freshness
• Retrieval evaluation
• Structured knowledge graphs

• Prompt evaluation
• Agent evaluation
• Tool-call evaluation
• Trace-based evaluation
• Retrieval quality evaluation
• Rubric-based scoring
• Failure analysis
• Hallucination detection
• Regression testing
• Feedback loops
• Self-improving workflows
• Cost and latency monitoring

• n8n workflows
• Webhooks
• REST APIs
• Scheduled workflows
• Event-driven automation
• Support automation
• Research automation
• Content workflows
• Approval-based workflows
• Human-in-the-loop systems
• Monitoring and alerting workflows

I work with hosted APIs, local inference stacks, and model routing layers for agents that need privacy, speed, offline workflows, fallback routing, or full control over behavior.

Local inference and agent runtime

I'm open to collaborating on:

• LLM agents and agentic workflows
• Local AI agents
• AI-ready knowledge systems
• AI training and model evaluation
• RAG and context engineering
• n8n AI automation
• MCP and tool-using agents
• AI observability and evaluation tools
• Voice agents and multimodal workflows
• Open-source AI infrastructure

• GitHub: github.com/oxbshw
• X: @Sayedevv
• Bluesky: sayedev.bsky.social

Building AI systems that connect language models with tools, knowledge, evaluation, automation, and real work.