AlphaAgents for Stock Evaluation

Personal learning project to implement the debate-driven multi-agent framework from BlackRock’s Aug 2025 paper for stock evaluation.

Aug 2025 – PresentSystems Engineer

Highlights

  • Implemented domain-specific agents that emit structured reports with reasoning, metrics, and confidence scoring.
  • Built a debate engine that streams multi-round critiques over SSE while logging every turn to reasoning_trace.jsonl.
  • Developed a coordinator module that applies majority voting and deterministic tie-breakers to synthesize consensus.
  • Visualized cumulative returns, rolling Sharpe, and drawdowns with Matplotlib dashboards backed by mock backtests.

Outcomes

  • Delivered an explainable baseline mirroring BlackRock’s research and ready for reinforcement-learning extensions.

Let's connect

I'm exploring AI agents, market-tech systems, and music-driven interfaces. Message me on LinkedIn if you want to collaborate or chat about tech.

Message on LinkedIn