【Company Profile】
HIGGS HK TECHNOLOGY LIMITED is a leading provider of AI-driven solutions for the financial trading industry. We are committed to building high-performance, AI-powered infrastructure for China’s financial sector, helping trading teams efficiently manage models and datasets within regulatory frameworks, conduct deeper market research, and make more informed trading decisions. Our core values are “Uphold Integrity, Innovate with Insight, Maintain Humility, and Act Altruistically.” We have international development teams located in Kuala Lumpur and Ho Chi Minh City.
【Position Overview】
We are seeking an engineer highly proficient in DQN (Deep Q-Network) and advanced deep reinforcement learning techniques to support the development and deployment of High-Frequency Trading (HFT) strategies. This role demands a deep understanding of Q-learning and its enhancements (e.g., Double DQN, Q-teacher, KL regularization) and the ability to handle large-scale sequential data, complex market volatility, and real-time performance requirements in HFT scenarios. You will be responsible for designing microsecond-to-second level RL trading agents, exploring Hierarchical RL with multiple time scales, and continuously optimizing trading returns and risk management.
【Key Responsibilities】
DQN Model Development & Optimization
Design and implement high-frequency trading strategies based on DDQN/Double Q-Learning, connecting to second-by-second or millisecond-level market data.
Integrate Q-teacher (dynamic programming / optimal value supervision) with classic TD updates to accelerate convergence and improve training stability.
Employ KL regularization, prioritized experience replay, and other cutting-edge techniques to boost exploration efficiency and reduce training time on large-scale, high-frequency data.
Trading Environment Construction & Reinforcement Learning Experiments
Build a high-fidelity simulation environment closely approximating real markets (including multi-level LOB, live order matching, etc.), and design a robust market order execution logic.
Develop hierarchical / multi-time-scale RL workflows (e.g., second-level low-level strategies, minute-level routing decisions) that leverage market microstructure insights.
Create a scalable large-scale parallel experimentation framework, simulating hundreds of millions of time-series data points in batch for training and evaluation.
Data Preprocessing & Feature Engineering
Collaborate with data engineers and quantitative researchers to acquire and clean multi-source market data (e.g., Limit Order Book (LOB), OHLC, technical indicators).
Utilize Talib-based indicators (MACD, order imbalance, VWAP, etc.) alongside custom features to enhance DQN’s perception of market price dynamics.
Apply techniques such as time-series differencing, denoising, segmentation to accurately model bull, bear, and sideways markets.
Model Evaluation & Risk Management
Develop comprehensive performance and risk metrics (e.g., annualized return, Sharpe ratio, max drawdown) for multi-dimensional evaluation of DQN-based strategies.
Implement multi-strategy routing/pooling (combining multiple sub-strategies that excel in different market segments) to mitigate single-strategy failure.
Examine model robustness under extreme market conditions (large volatility, network latency, etc.) and propose improvements or risk mitigation measures.
Continuous Optimization & Cross-Team Collaboration
Work closely with quantitative teams to understand business needs, align model predictions, and provide explainable strategy suggestions.
Draft and maintain relevant technical documentation and deployment scripts, offering training and support to internal teams and external partners.
Stay current with research trends in Hierarchical Reinforcement Learning, Dynamic Programming, Financial Time Series, and integrate new findings into production trading systems.
【Qualifications】
Education & Background
Bachelor’s degree or above in Computer Science, Statistics, Mathematics, Financial Engineering, or related fields.
Preferably with project experience in high-frequency trading, quantitative trading, or financial risk management.
Technical & Modeling Expertise
Deep understanding of Q-learning and its variants (DQN, Double DQN, DDQN), including key techniques such as value function approximation, experience replay, and target networks.
Proficient in Python and data analysis libraries (NumPy, Pandas, PyTorch/TensorFlow), capable of independently building deep networks, training, and tuning models.
Familiarity with Hierarchical RL, including methods to incorporate multi-time-scale information (e.g., second-level, minute-level) into model frameworks.
Knowledgeable about LOB (Limit Order Book) and OHLC financial microstructure data, with a solid foundation in dynamic programming and optimal strategy estimation.
Able to analyze standard financial performance metrics (return, max drawdown, volatility, Sharpe ratio, etc.) for multi-dimensional assessment of high-frequency strategies.
Programming & Tools
Proficient in C++ and Python, familiar with Git, Docker, and related DevOps tools; experience in distributed or parallel training is a plus.
Skilled in writing high-quality, reproducible code for both experimental and production environments, with expertise in unit testing and performance optimization.
Familiar with large-scale time-series processing and real-time data ingestion; practical experience in building high-throughput data pipelines using Redis, Flink, Kafka, or common databases (SQL/NoSQL).
General Skills
Strong quantitative background and analytical thinking to address complex model or data challenges.
Excellent communication and teamwork abilities, collaborating effectively with quant, data engineering, and risk management teams.
Active interest in academic and industry advances, staying up-to-date on techniques such as dynamic programming and hierarchical strategy routing within high-frequency trading.
【Compensation & Benefits】
Competitive salary and year-end performance bonuses.
Flat organizational structure with a dynamic, supportive tech team culture.
Flexible work arrangements, with up to 100% remote options; overseas work allowances available annually.
Multiple company-sponsored overseas trips each year and a variety of team-building activities.
Broad career development opportunities, applying cutting-edge reinforcement learning and quantitative finance technologies.
【Work Locations】
Kuala Lumpur, Malaysia & Ho Chi Minh City, Vietnam