Hongxu Lai | Adrian

定位 定位 Positioning

我构建会接触真实资金的交易系统。我从 Python 开始，但正在转向用 Modern C++ 构建低延迟系统。我关注让系统在生产环境中可靠的部分：订单状态、执行、持仓更新、风险控制和恢复。 我构建会接触真實资金的交易系统。我从 Python 开始，但正在转向用 Modern C++ 构建低延迟系统。我关注让系统在生产环境中可靠的部分：订单状态、执行、持仓更新、风险控制和恢复。 I build trading systems that touch real money. I started with Python, but I am moving toward low-latency systems in Modern C++. I care about the parts that make a system reliable in production: order state, execution, position updates, risk controls, and recovery.

• Modern C++：所有权、值语义、模板、并发、测试和基于 CMake 的构建。 Modern C++：所有权、值语义、模板、并发、测试和基于 CMake 的构建。 Modern C++: ownership, value semantics, templates, concurrency, testing, and CMake-based builds.
• 交易系统：市场微观结构、订单簿、执行质量、风险边界和账户状态。 交易系统：市场微观结构、订单簿、执行质量、风险边界和账户状态。 Trading systems: market microstructure, order books, execution quality, risk boundaries, and account state.
• 操作系统与网络：进程、调度、内存、套接字、背压和故障隔离。 操作系统与网络：进程、调度、内存、套接字、背压和故障隔离。 Operating systems and networking: processes, scheduling, memory, sockets, backpressure, and failure isolation.
• 量化研究基础设施：数据卫生、可回放实验、度量指标，以及把假设转化为实现。 量化研究基础设施：數據卫生、可回放实验、度量指标，以及把假设转化为实现。 Quantitative research infrastructure: data hygiene, replayable experiments, metrics, and turning hypotheses into implementations.

当前项目 当前項目 Current Projects

• jev-quant用于连接策略决策、行情数据、订单管理、持仓跟踪和风险检查的 C++20 交易系统。交易引擎通过有界队列传递类型化消息，由独立组件处理行情更新、订单回报、Jev 决策和交易所请求。Jev 是一个提供带有概率和置信度的类型化交易决策的 AI 模型。专用 worker 将这些决策发送给交易引擎，再由 C++ 处理订单细节、风险检查和状态变化。 用于连接策略决策、行情數據、订单管理、持仓跟踪和风险检查的 C++20 交易系统。交易引擎通过有界队列传递类型化消息，由独立组件处理行情更新、订单回报、Jev 决策和交易所请求。Jev 是一个提供带有概率和置信度的类型化交易决策的 AI 模型。专用 worker 将这些决策发送给交易引擎，再由 C++ 处理订单细节、风险检查和状态变化。 A C++20 trading system for connecting strategy decisions to market data, order management, position tracking, and risk checks. Its trade engine passes typed messages through bounded queues, while separate components handle market updates, order responses, Jev decisions, and venue requests. Jev is an AI model that provides typed trading decisions with probability and confidence values. A dedicated worker sends these decisions to the trade engine, where C++ handles order details, risk checks, and state changes.
• simex一个带有上期所风格 RB 期货配置的 C++ 模拟交易所。它模拟虚拟时间、交易时段、交易日切换、价格限制，以及集合竞价和连续竞价。订单按价格优先和时间优先撮合，模拟器发布带序号的行情增量和独立快照。JSON 回放轨迹和最终 SHA-256 状态哈希让每次运行都可以复现。 一个带有上期所风格 RB 期货配置的 C++ 模拟交易所。它模拟虚拟時間、交易时段、交易日切換、价格限制，以及集合竞价和连续竞价。订单按价格优先和時間优先撮合，模拟器发布带序号的行情增量和独立快照。JSON 回放轨迹和最终 SHA-256 状态哈希让每次运行都可以复现。 A C++ simulated exchange with an SHFE-style RB futures profile. It models virtual time, trading-session phases, trading-day rollover, price limits, and both auction and continuous matching. Orders are matched by price and time priority, while the simulator publishes sequenced market-data updates and independent snapshots. JSON replay traces and a final SHA-256 state hash make each run reproducible.

实习经历 实习经历 Internships

• Guosheng Securities

  电子行业组实习。我收集行业数据和公司新闻，完成分析，并准备工业富联首次覆盖报告的初稿。团队审阅后发布了报告。这段经历让我了解如何从公司信息、行业证据和清晰的报告结构中建立投资判断。 电子行业组实习。我收集行业數據和公司新闻，完成分析，并準備工业富联首次覆盖报告的初稿。团队审阅后发布了报告。这段经历让我了解如何从公司信息、行业证据和清晰的报告结构中建立投资判断。 Equity research intern on the electronics team. I collected industry data and company news, developed the analysis, and prepared the first draft of an initiation report on Foxconn Industrial Internet. The team reviewed and published the report. The work taught me how an investment view is built from company information, industry evidence, and a clear report structure.

• Hunan Wanjun Capital Management

  我实现并维护了一个结合 TD 信号的 MACD 指数期权策略，并参与多腿期权策略的实现。之后，我将这些策略及其配套服务迁移到共享交易基础设施上。迁移后，多个策略共用交易所连接、订单成交持仓记录和统一配置。 我实现并维护了一个结合 TD 信号的 MACD 指数期权策略，并参与多腿期权策略的实现。之后，我将这些策略及其配套服务迁移到共享交易基础设施上。迁移后，多个策略共用交易所连接、订单成交持仓记录和统一配置。 I implemented and maintained a MACD-based index options strategy with TD signals and supported a multi-leg options strategy. I later moved these strategies and their supporting services onto a shared trading infrastructure. The migration introduced common exchange connections, centralised order, fill, and position records, and unified configuration across strategies.

精选作品 精选作品 Selected Work

下面这些项目代表我想持续做的那种工作：有清晰运行故事的软件，让正确性可见，并把用户的决策与底层状态变迁连接起来。 下面这些項目代表我想持续做的那种工作：有清晰运行故事的軟件，让正确性可见，并把用户的决策与底层状态变迁连接起来。 The projects below show the shape of work I want to keep doing: software that has a clear operational story, makes correctness visible, and connects a user's decision to the underlying state transitions.

• AlphaFlow

  一个智能体辅助的研究工作流，把问题转化为结构化证据：并发有边界、重试显式化，产出可审阅而不只是被生成。 一个智能体辅助的研究工作流，把問題转化为结构化证据：并发有边界、重试显式化，产出可审阅而不只是被生成。 An agent-assisted research workflow that turns a question into structured evidence, with bounded concurrency, explicit retries, and outputs that can be reviewed rather than merely generated.

  LangGraph · Next.js · PostgreSQL · Redis

• 订单与持仓状态管理 订单与持仓状态管理 Order and position state management

  面向实盘期权交易的事件驱动对账，围绕账户真实状态设计：重启安全的快照，以及订单更新与已结算持仓之间的区别。 面向实盘期权交易的事件驱动对账，围绕账户真實状态设计：重启安全的快照，以及订单更新与已结算持仓之间的区别。 Event-driven reconciliation for live options trading, designed around account truth, restart-safe snapshots, and the difference between an order update and a position that is actually settled.

  VeighNa · CTP · MySQL · Docker