Our PIPE Phase II was approved by the São Paulo Research Foundation — Why it matters

I’m glad to share that the São Paulo Research Foundation approved Phase II of our project, Deep Reinforcement Learning for Portfolio Management.

Their support goes beyond funding: approval requires peer review and adherence to a scientific method at every step, ensuring rigorous experimentation, validation, and transparency.

DRL in plain words

Instead of merely forecasting prices, Deep Reinforcement Learning teaches an agent to make sequential allocation decisions (buy/sell/hold), learning from rewards and penalties over time. The goal is to learn policies that balance return and risk dynamically — crucial for portfolios that must adapt to regime shifts.

What we delivered in Phase I

• Built and validated multiple DRL models and quantitative prototypes.

• Launched four strategies, each with a distinct profile:

  • All Caps (broad equities)

  • Energy (oil & gas)

  • Semiconductors

  • Defensive Yield (income/dividends).

• Integrated with a Broker (Interactive Brokers), so everything is automated, now

• Created an app for portfolio management (Tekton Alpha - www.tektonalpha.com)

• Created an app for managing AI processes in multiple virtual machines (Opt Viewer)

  
  

What Phase II will deliver

• Broader asset classes and data granularities (daily and intraday).

• Stronger DRL (and ensemble) policies with robust out-of-sample testing.

• A production-grade research platform to accelerate study, validation, and publication.

  
  

Why advisors and investors should care

• Context-aware allocation (no rigid static assumptions).

• Discipline and scale, limiting behavioral bias.

• Customization to mandates (growth, income, volatility targets).

  
  

#AI #DRL #QuantFinance #Research #Innovation #PIPE #SaoPauloResearchFoundation