English | 简体中文

I am a Big Data Engineer specializing in quantitative research infrastructure, data pipelines, and AI-driven trading systems.

Dedicated to building practical, end-to-end quantitative infrastructures that seamlessly integrate the entire research and trading lifecycle.

🎯 Ultimate Vision: Powered by Autonomous Reasoning & Control (ARC) principles, my ultimate goal is to engineer self-evolving, autonomous agent systems capable of independent exploration, continuous strategy discovery, and adaptive execution within highly complex financial environments.

[ Market Data Pipeline ] ──> [ Factor Mining & Feature Engineering ] ──> [ Strategy & Alpha Research ]
                                                                                   │
[ Automated Execution  ] <── [ Paper Trading & Signal Audit        ] <── [ Backtesting & Simulation  ]

• Real-Time Streaming: Apache Flink · Kafka · Tick/Bar Streaming · Flink SQL · Large-State Checkpoint Tuning
• PB-Scale DWH & Batch Processing: Apache Spark · Hive · Hadoop · ODS ➔ DWD ➔ DWS Layered DWH
• Workflow Scheduling & Governance: Airflow · In-House Distributed Scheduler · 1,000+ Job DAG Governance · SLA Monitoring & Recovery
• Storage, Analytics & Execution: ClickHouse · HBase · PostgreSQL · MySQL · Execution Gateways · Risk Controls

I am actively developing and maintaining a suite of quantitative research tools and production-grade software:

A production-grade, governed quantitative research and strategy incubation Agent Runtime powered by the universal ARC (Autonomous Research Core) engine:

• MCTS & MAP-Elites Search Engine: Combines Monte Carlo Tree Search over strategy code ASTs with Quality-Diversity grid archiving to explore high-dimensional strategy spaces without premature convergence.
• Adversarial Red-Teaming: Blue Team quant agents formulate Alpha hypotheses while Red Team agents stress-test for black swan shocks, liquidity traps, and stop-loss vulnerabilities.
• Multi-Regime Causal Attribution & Reflexion: Deconstructs performance across market regimes (trending, volatile, range-bound) and distills structured negative constraints for continuous prompt feedback.
• Voyager-Style Skill Distillation & Paper Trading: Automatically distills validated code sub-functions into an immutable skill library, deploying robust candidate strategies to paper trading environments zero-touch.

A universal autonomous program synthesis and AGI reasoning engine designed for the full ARC-AGI benchmark suite (ARC-AGI-1, 2, and 3 / ARC Prize 2026):

• Universal ARC Benchmark Suite: Standardized task models (ARCTask) and automated dataset loaders supporting ARC-AGI-1, 2, and 3.
• Parallel MCTS Solver Engine: Multi-threaded AST search engine (HyperARCParallelMCTSEngine) executing parallel program mutation rollouts over 2D spatial grid transformations.
• 2D Grid DSL Primitives: Rich domain-specific primitives for spatial operations (rotate_90, flip_horizontal, replace_color, crop_bounding_box).
• Self-Healing Harness & Exact Matching: Scaffolding with error recovery (HyperARCHarness) that enforces 100% pixel-exact matching on training grid examples before predicting unseen test grids.

A-share research and monitoring platform covering real-time market data, AI stock evaluation, factor research, strategy development, and simulation trading.

Open-source quantitative research workbench focused on real market data, backtesting, paper trading, signal audit, and risk-first strategy development.

Self-evolving A-share stock selection system combining Kronos K-line forecasting, Hermes Agent loops, and a three-pool funnel workflow.

A WeChat mini program I'm continuously operating and improving, with promotion through WeChat 搜一搜.

A fishing-focused tide and weather mini program I'm continuously operating and improving, with promotion through WeChat 搜一搜.