About Highlandy Investment LLC
Highlandy Investment LLC is a proprietary AI quantitative investment firm founded in Dec. 2024. We design, backtest, and deploy high-performance algorithmic trading strategies using generative AI, deep reinforcement learning, graph neural networks, and real-time inference infrastructure.
Our mission is to merge rigorous statistical research with state-of-the-art Al prediction algorithms and computational engineering to deliver robust alpha generation across US stock market.
Our current 10-month 30% gain (15% over SPY as 09/07/2026 and sharpe ratio of 1.6) indicates we can move on to the next phase. We expect to expand our model sizes by 10x and develop more complex reinforcement learning models.
The Quantitative Investment Approach
In ever-changing global markets, navigating bull and bear cycles while generating stable profits remains the ultimate objective for investors. Highlandy Investment LLC—a quantitative investment firm rooted in Silicon Valley—is leveraging AI to consistently outperform benchmark targets.
We frequently reflect on how to achieve stable wealth preservation and appreciation over the long run. To avoid artificial volatility and allow asset values to be fully priced by the market, we choose mega-cap stocks as our primary targets. Simultaneously, we emphasize evaluating the true value of equities, accurately timing low-price entry points and position sizing to capture long-term growth driven by real-world corporate value creation. We also respect irrational market fluctuations, purely extracting underlying patterns from massive datasets to capture fleeting, short-term opportunities.
To maximize AI efficiency while minimizing hardware dependence, we reference leading models such as Gemini 3.1 Pro, Muse Spark, Claude 4.5, and o4-mini to reconstruct a lightweight stock quantitative trading platform from the ground up, making intelligent decision-making more efficient, stable, and actionable.
Led by Chao Cheng (a seasoned Silicon Valley AI/Machine Learning and Big Data Scientist), the team created a customized AI investment research platform powered by over a hundred models. This platform constructs a multi-scenario strategy matrix: ranging from left-side conservative strategies ("buy low, sell high") to right-side offensive strategies ("buy high, sell higher"), as well as aggressive call option strategies designed to capture index movements, precisely adapting to different risk preferences and market environments. Crucially, based on deep backtesting over 30 years of market data, the team built a defensive mechanism against unpredictable risks and black swan events, laying a solid firewall for asset safety.
Even more groundbreaking, HighLandy Investment LLC has achieved fully automated stock selection and execution, eliminating the uncertainties associated with manual intervention. Traditional investing relies heavily on traders' subjective judgments, leading to greed during rallies and fear during drops, where emotional volatility often yields irreversible losses. In contrast, our AI system operates 24/7, strictly executing according to preset models and risk management rules. From stock selection, timing, and order placement to position adjustments, profit-taking, and stop-loss execution, every step is automated by algorithms—completely devoid of emotion, speculation, or external interference. This fundamentally eliminates human errors driven by greed, fear, or hesitation, entrusting decision-making and execution entirely to data and machines to maximize the algorithm's stability, efficiency, and discipline.
Quantitative Logic Breakdown: Closed-Loop Verification from Signal to Return
The core of quantitative investing lies in replacing emotional decisions with reproducible rules, using data-driven disciplined trading to achieve long-term, stable excess returns. Our three core charts illustrate the complete logic from signal generation and capital management to return realization.
1. Trading Signals: Disciplined Execution of Left-Side Value Positioning
In the first asset trading chart, the AI model's precise timing for "buying low" is clear:
Signal Triggering: Gray plus signs represent AI buy signals, concentrated in lower price ranges (e.g., the second half of 2024 and mid-2025) where valuations were at historical lows, matching the quantitative criteria for "value troughs."
Cost Control: The green stepped line shows the dynamic average holding cost. Rather than entering a full position all at once, positions are built incrementally as signals trigger. This classic left-side batch position-building strategy avoids single-point timing errors and lowers average cost to create a safety margin for subsequent rallies.
Market Adaptation: After the stock price broke through key resistance levels at the end of 2025, the average cost remained stable, showing that the model stopped chasing prices once the trend was confirmed and shifted to collecting holding returns—demonstrating quantitative discipline.
2. Capital & Returns: Maximizing Capital Efficiency Under Quantitative Risk Control
The second account monitoring chart illustrates the dynamic balance among return, risk, and capital management:
Return Realization: The black stepped line (cumulative returns) displays a characteristic "stepped ascent." Each step corresponds to a full cycle of position building, price appreciation, and position reduction, showing that the model captures upward moves while using profit-taking mechanisms to convert unrealized gains into realized profits.
Unrealized Profit & Cash Management: The purple curve (unrealized profit) rises rapidly during uptrends and pulls back during corrections. Paired with periodic shifts in the blue curve (cash utilization), it executes dynamic capital management ("reduce position on rallies, add on pullbacks"), ensuring capital remains focused on high value-to-cost assets.
Risk Isolation Mechanism: The cash curve remains low during periods of sharp market volatility (e.g., 2022 and early 2024), indicating that the model actively dodges systemic risk by reducing position sizes and holding cash—a direct application of our "black swan defense mechanism."
Sharpe Ratio control: We also apply reinforcement learning (RL) to dynamic portfolio management with sharpe ratio control as target.
3. Benchmark Comparison: Verification of Outperformance Across Cycles
The third return comparison chart serves as the final validation of strategy effectiveness.
significant uptrend potentials and Excess Returns: From November 2025 through September 2026, the strategy return (blue curve) first peaked above 20% in January 2026, whereas the S&P 500 (green curve) peaked near 3% over the same period, yielding an excess return of nearly 17 percentage points. As of the time of writing this article (09/07/2026), our strategy continues to maintain an additional return of 15% over SPY.
Downside Resilience and return stability: During the market pullback in March 2026, when the S&P 500 experienced a notable drawdown, the strategy return also saw similar fluctuations before rebounding quickly, demonstrating resilience during extreme market conditions. We optimized our models in terms of position sizing, entry/exit timing and hedging strategies. We can observe obvious volatility improvement over SPY since early Jun 2026.
Indicator tools Cross US Market
We show two SP500 indicators as demonstration examples of our proprietary AI Quant tools as of 04/13/2026;
U.S. Market Indicator Applications
Dynamic Cash Allocation: Reduces cash drag during low or negative market indicator regimes; increases cash reserves when market indicators reach elevated levels.
Tactical Options Execution: Triggers SPY call option entries upon peak indicator signals to capture asymmetric momentum upside.
Look into the future
Amid increasingly noisy markets, Highlandy investment LLC utilizes AI quantitative strategies to redefine traditional investing boundaries, proving the power of technology-enabled investment through data-backed performance. Whether aiming for steady capital preservation or seeking excess returns, Highlandy Investment LLC offers tailored solutions to help assets navigate market cycles for long-term growth.
Since our initial lightweight model went live and began serving investors in early November 2025, it has achieved visible investment return advantages over SP&500. This has further strengthened our resolve to increase investment in two directions—better fine-tuning strategies and larger foundational models—striving to deliver even more substantial returns for our clients. Although more complex fine-tuning strategies and larger models entail greater hardware investments, both investors and the company itself stand to benefit from larger and more robust investment returns.
Leadership
Chao Cheng, PhD
Founder & Lead Quantitative Strategist
https://www.linkedin.com/in/chao-cheng-b229943/
Chao Cheng leads quantitative strategy design, machine learning architecture, and system infrastructure at Highlandy Investment LLC.
Education: PhD in Electrical and Computer Engineering from the University of Minnesota.
Background: Extensive engineering and research background spanning quantitative trading algorithms, computer architecture, large-scale deep learning models, and system-level optimization across leading tech and research institutions (Highlandy Investment LLC, TikTok, Alibaba, Qualcomm).
Core Tech. Pillars
1. Advanced Machine Learning & Reinforcement Learning
Custom PyTorch and execution pipelines designed for real-time model training and inference.
Signal discovery leveraging Generative AI, Graph Neural Networks (GNNs) and deep reinforcement learning (GRPO/PPO) for dynamic market modeling.
2. Institutional Quantitative Research
Systematic strategy development focused on market inefficiencies, RSI divergence, statistical arbitrage, and risk-controlled portfolio allocation.
Mathematical validation and high-frequency data analysis to maintain strategy resilience across changing market regimes.
3. Low-Latency Compute Architecture
Custom-engineered high-density multi-GPU server clusters powered by specialized liquid cooling systems.
High-throughput data ingestion and low-latency execution engines built for performance and maximum uptime.
Contact Us
Interested in strategic partnerships or quantitative research collaboration?
Location: Santa Clara, California
Email: chao@highlandyinvest.com
Website: https://www.highlandyinvest.com