AI Infrastructure Engineer

Liang Li

I help teams turn LLM demos into production AI agents — reliable agent, RAG, and MCP systems that connect to your real tools, APIs, and data, so an agent can actually take action instead of just chatting. Built end-to-end in Go & Rust.

OPEN TO FREELANCE — agents · RAG · MCP Hire me → GitHub
Location
Chengdu, CN · Remote
Stack
Go · Rust · TS/React
Focus
Agents · RAG · MCP · Local LLM
Status
● Available for contract
01Services — what you can hire me for

Agent System Architecture

For
Teams with an LLM prototype that needs to run reliably in production.
Get
Agent loop, tool & skill integration, memory, evals, and guardrails.
Time
Typically 2–6 weeks.

RAG & Knowledge Graphs

For
Teams that need agents to answer accurately over their own data.
Get
Ingestion, vector + hybrid retrieval, GraphRAG — embedded or hosted.
Time
Typically 2–4 weeks.

MCP Tool Integration

For
Teams whose agents need to take real actions in real systems.
Get
Custom MCP servers wiring up internal tools, APIs, and databases.
Time
Typically 1–3 weeks.
02Agent Frameworks & Orchestration
harness-rs RUST
A production-ready Rust foundation for custom agents — so you build on a solid base instead of from scratch. ReAct loop, pluggable tools & skills, cross-session memory, scheduler, sandbox, and MCP client/server. Published on crates.io.
agent-go GO
Drop AI agents straight into your existing Go services — teams, tasks, memory, MCP, and tool calling, with fewer external dependencies to deploy.
roma GO
Coordinate several agents on one task — a runtime orchestrator for multi-agent systems.
03Memory · Knowledge Graphs · RAG
cortexdb ★ flagship GO
Give agents reliable retrieval without standing up a separate database. A pure-Go, single-file AI memory & knowledge-graph engine: vector + hybrid (BM25/FTS5) search, GraphRAG, one-pass structured-data import — zero external services, fully embedded.
askdoc GO
Point an agent at your own documents and get grounded answers — a working RAG pipeline for internal docs.
04MCP Servers — connect anything to an agent
mcp-swagger-server GO
Connect an existing API to an agent without writing glue code — turn any Swagger / OpenAPI spec into ready-to-use agent tools, cutting integration setup time.
mcp-websearch-server GO
Give agents live, multi-engine web search with content extraction — exposed as MCP tools.
mcp-sqlite-server GO
Let an agent safely query a SQLite database over MCP. (+ mcp-snapshot-server)
05Local-LLM & Infrastructure Tooling
ollama-go GO
Run local models from your Go code — a Go client library for Ollama, for teams that need on-prem or private LLMs.
lmstudio-go GO
A Go client for LM Studio: chat, embeddings, tool calling, model management.
gosible · dispatch GO
Infrastructure & multi-server automation in Go. (+ ollama-queue: a task queue for Ollama)
06FAQ
What does Liang Li do?

I help teams turn LLM demos into production AI agents — building agent systems, RAG and knowledge-graph retrieval, and MCP integrations that connect agents to your real tools, APIs, and databases. Primarily in Go and Rust.

Can you take our AI prototype from demo to production?

Yes — that's the core of the work. I take an LLM prototype and turn it into a reliable system that runs unattended inside your real environment, with evals, guardrails, documentation, and a clear deployment path.

Can you work with our existing codebase?

Yes. I'm comfortable joining an existing codebase and integrating agent, RAG, and MCP capabilities into systems you already run — not only greenfield projects.

Do you build custom MCP tools for internal systems?

Yes. I build custom MCP servers that let agents take real actions — connecting internal tools, APIs, and databases so an agent can do useful work, not just chat.

Do you work async with US / EU teams?

Yes. I'm remote-first and async-friendly, based in Chengdu (UTC+8), and work with US and EU teams through clear written communication and documentation. Contract and freelance engagements — reach me at ll_faw@hotmail.com.

What are the main projects?

cortexdb (a pure-Go AI memory + knowledge-graph engine with vector/hybrid search and GraphRAG), harness-rs (a Rust agent framework on crates.io), agent-go (a Go AI agent SDK), and a suite of MCP servers for Swagger/OpenAPI, web search, and SQLite.

I take an LLM from “demo” to running unattended inside your systems.