thomassss
Ai Agent Developer
AI Agent / LLM Systems Engineer focused on production trading & infra use‑cases.
I design and build long‑running, tool‑using agent systems that sit on top of real markets, real infrastructure and real incident workflows.
I design and build long‑running, tool‑using agent systems that sit on top of real markets, real infrastructure and real incident workflows.
Most recently on the GMX‑Solana core team, I led the design of two agentic systems:
- GMX‑Solana AI Trading Copilot – a strategy and execution assistant built on OpenClaw + LangGraph + LangChain. I wrapped our keeper, Subsquid‑based market stats and Solana SDK as tools so the agent can aggregate on‑chain data, run risk checks, and generate multiple executable strategies (leverage, sizing, TP/SL) with corresponding unsigned transactions / order payloads. The Copilot is integrated directly into the trading UI, while keeping a strict security boundary: users always sign with their own wallets; neither backend nor OpenClaw ever touch private keys.
- Keeper Ops AI Copilot – an internal SRE assistant that performs root‑cause analysis across GMSOL Keeper, indexers and off‑chain services. Using OpenClaw as an incident orchestration layer plus LangChain + LlamaIndex, it reacts to latency spikes, queue backlogs or error surges by querying Prometheus/log stores, inspecting keeper configs and searching the codebase, then surfaces likely causes and concrete remediation steps. We expose this via an SRE console with AI‑generated RCA reports and actionable runbooks, turning incidents into a structured knowledge base and reducing MTTR.
My focus is on agent orchestration, tool design, and reliability rather than just calling an LLM: defining task lifecycles, tool schemas, guardrails, and audit trails; instrumenting agents with tracing and logging for post‑mortems; and integrating them cleanly into existing backends and frontends.
I come from a strong systems and trading background (Rust/Python/Solidity, low‑latency execution, keeper/arb engines, Solana/EVM infra), which helps me reason about risk, performance and failure modes when agents are allowed to operate on live systems. I’m looking for roles where I can own AI agent architecture end‑to‑end—from problem framing and tool design to orchestration, observability and production rollout—especially around trading, risk, DevOps/SRE, or other high‑stakes automation domains.
Experience: 4 years
Yearly salary: $30,000
Hourly rate: $20
Nationality: 🇭🇰 Hong Kong
Residency: 🇭🇰 Hong Kong
Experience
AI Agent Developer
GMX Limited 2025 - 2026
GMX-Solana is an on-chain perpetual futures and swap protocol deployed on Solana. It provides unified liquidity for multiple markets via GLV (Global Liquidity Vault) and GM pools, and has processed several billions of dollars in cumulative trading volume on mainnet. 1.GMX‑Solana AI Trading Copilot (Intelligent Strategy & Execution Assistant) 2.Keeper Ops AI Copilot (Operations & Root Cause Analysis) Tech stack: Rust, Anchor, Solana, Chainlink, Pyth, multi-market perpetual protocol design, AMM/liquidity pool models, GLV/GM liquidity architecture, off-chain keeper / order execution engine, OpenClaw, LangChain, LangGraph, LlamaIndex, LLM‑based agent system design.
Skills
ai
blockchain-engineer
cosmos
rust
english