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.

• 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 building and exploring practical systems around autonomous reasoning, quantitative strategy discovery, and real-world product engineering:

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.

Competitions: ARC-AGI-2 (Kaggle) · ARC-AGI-3 (Kaggle) · ARC Prize 2026

An autonomous program synthesis and AGI reasoning research system designed for the full ARC-AGI benchmark suite (ARC-AGI-1/2 program synthesis and ARC-AGI-3 interactive-Agent experiments):

• Grid-Transformation DSL & MCTS Solver: 2D spatial primitives (rotate_90, flip_horizontal, replace_color, crop_bounding_box) combined with multi-threaded AST search (HyperARCParallelMCTSEngine).
• Exact-Match Validation: Self-healing harness scaffolding (HyperARCHarness) enforcing 100% pixel-exact matching on training grid examples before predicting unseen test grids.
• Visual-State Abstraction & Trajectory Evaluation: State abstraction, action history backtracking, skill routing, and trajectory-based evaluation.

• Alpha — Self-evolving A-share stock selection system combining Kronos K-line forecasting models, Hermes Agent loops, and a three-pool funnel workflow.
• StockPro — Real-time A-share research and monitoring platform covering real-time market data, AI stock evaluation, factor research, strategy development, and simulation trading.
• QuantBase — Open-source quantitative research workbench focused on real market data, backtesting, paper trading, signal audit, and risk-first strategy development.

A food-ingredient analysis and health-literacy mini program that I continuously operate and improve, covering data organization, product iteration, and promotion through WeChat Search.

This product is continuously operated and iterated—not a one-off demo.