ProjectAlpha

Project Alpha Project Alpha is a grounded AI research platform for professional portfolio managers. Built for boutique asset managers, family offices, and hedge funds managing less than $500 million in AUM, it brings investment research, proprietary data, and AI-driven strategy development into a single reproducible workspace. Unlike conventional AI assistants, Project Alpha is engineered around one principle: the AI never invents a number. Rather than generating quantitative outputs itself, the language model serves only as an orchestration layer. Every financial metric, statistic, and backtest result is produced by deterministic analytical engines through typed tool calls. Before any result reaches the interface, an independent validation layer verifies that every numerical value is backed by an authenticated computation. Unsupported values are automatically blocked. Every figure is fully traceable. Clicking any number reveals its complete provenance, including the originating dataset and version, as-of timestamp, analytical code version, execution seed, and supporting computation—making every result transparent, reproducible, and auditable. Project Alpha is designed for experienced investment professionals rather than programmers. Portfolio managers and founders can conduct sophisticated research through an intuitive desktop interface, while analysts and quantitative researchers can extend workflows through code-enabled surfaces. The platform complements existing institutional tools such as Bloomberg rather than replacing them, serving as the firm's daily AI-powered research environment. Today, Project Alpha includes three core experiences: Sandbox — A chart-first research environment for exploring fully reproducible backtests, featuring interactive equity curves, benchmark comparisons, drawdown analysis, rolling Sharpe ratios, and headline performance metrics. Copilot — A compact command palette that answers questions about any research run using grounded, cited responses. Every answer includes clickable citations, provenance popovers, and clear verification indicators showing whether information has been verified or blocked by the validation engine. Improvement Workspace — An AI-assisted strategy iteration environment that organizes research into an evolution tree while continuously monitoring overfitting risk through institutional-grade validation metrics, including Deflated Sharpe Ratio, Probability of Backtest Overfitting (PBO), walk-forward testing, and locked holdout evaluation. Project Alpha gives investment teams the speed and flexibility of modern AI while preserving the determinism, transparency, and reproducibility required for professional investment decision-making.