Atlas Investment Technologies, LLC
Systematic, machine learning driven trading across liquid U.S. ETFs.
Systematic, intraday, ML based trading focused on leveraged index and sector ETFs.
Atlas Investment Technologies employs a fully systematic, machine learning driven trading strategy focused on liquid U.S. exchange-traded funds, with an emphasis on leveraged index and sector ETFs. The system combines multi-horizon forecasting, intraday regime detection, and volatility-aware allocation to generate uncorrelated absolute returns across a range of market environments.
The objective is to deliver attractive risk adjusted returns with low correlation to major asset classes, while maintaining strict risk discipline throughout the trading day.
A proprietary ML framework produces short-horizon return, volatility, and regime forecasts using intraday data. Models incorporate cross-sectional relationships, time-of-day dynamics, and regime-conditioned expectations, allowing the portfolio to adapt quickly as conditions change.
Forecasts feed into a custom optimizer that balances expected returns against multiple risk measures, including CVaR, intraday volatility, and drawdown sensitivity. Diversification is enforced via entropy style constraints, with exposure scaled dynamically by forecast confidence and regime.
Portfolios rebalance intraday, avoid overnight leverage, and reduce exposure in adverse regimes. Execution is liquidity aware to limit impact and slippage. The strategy is fully systematic; human input is focused on model monitoring, risk oversight, and periodic retraining.
A disciplined, repeatable process from data ingestion to execution.
The Atlas Fund is a closed fund. To invest requires a formal invitation from the fund manager. There may be future financial products open to the general public; currently, all resources are private.
For eligible investors, the default fee schedule is a 2% management fee and a 20% performance fee (applied to positive returns above the benchmark). The minimum initial allocation is typically set at $200,000, and investors are strongly advised not to invest more than they are prepared to lose.
Withdrawals are generally processed at year-end. Investors submit requests via the application; subject to available liquidity, proceeds are distributed after Q4.
Funds may also be distributed to meet tax obligations or in the event of intentional liquidation. In extraordinary circumstances, the fund manager reserves the right to limit or defer withdrawals. Deposits can be made throughout the year but must be confirmed before they appear on an investor's dashboard.
Gregory is a former Google Machine Learning Engineer with a Master's degree in Computer Science (Machine Learning focus) and an undergraduate degree in Computer Science and Economics from Brown University. At Google he worked on consumer-facing ML systems at Google Cloud and on recommendation models at YouTube, including models that generated over $11 million in ARR and contributed to core metric improvements.
The strategy implemented in the Atlas Fund is an institutionalization of the systematic approach Gregory previously used for his own trading. The thesis is straightforward: leverage AI to amplify human macro and thematic intuition, while delegating intraday reaction and position sizing to models trained on millions of data points.
As fund manager, Gregory is responsible for supervising the research pipeline, monitoring model performance, managing risk parameters, and evolving the infrastructure that powers the Atlas Fund and its investor reporting platform.
Backtested and live results are monitored continuously; live performance is visible in the investor portal.
Performance data is inherently limited given the strategy's relatively recent deployment. In backtesting, under idealized assumptions and no frictions, the algorithm generated returns materially above benchmark levels across a range of regimes. Preliminary forward testing and live trading results have demonstrated the ability to generate positive returns even in challenging or bearish environments. From March 6, 2024 to April 30, 2024—during a broadly negative market window—the model returned approximately 10% while the S&P 500 declined more than 1%. While such results are not a guarantee of future performance, the consistency of the return profile and drawdown behavior across tests provides confidence that the core signal-generation and risk framework are behaving as designed.