Skip to main content
Powering Open Source
Democratising
Reinforcement Learning
Empower data scientists to solve sequential decision problems in hours not years. Fast, intuitive, and open — pi_optimal transforms reinforcement learning into a tool anyone can use to drive real-world impact.
Open Source Package
Prototype RL in 7 Lines of Code

Our package brings RL within reach for complex optimization problems. Whether it’s dynamic pricing or automated energy management, getting started is straightforward.

1. Install the Package

Install the package to get all the tools you need for building and training RL agents for complex business applications.

bash
$pip install pi_optimal

2. Load Your Dataset

Load historical data relevant to your problem. This example uses a building's energy consumption data.

python
1import pandas as pd
2
3# Load historical building control data
4historical_data = pd.read_csv("historical_building_control.csv")

3. Train the Agent

Configure and train the RL agent. Define actions, rewards, and let the agent learn the best strategies.

python
1import pi_optimal as po
2
3agent = po.Agent()
4agent.train(
5 data=historical_data,
6 action_columns=["adjust_lighting", "adjust_heating", "adjust_cooling"],
7 reward_column="energy_comfort_balance"
8)

4. Use the Agent

Apply the trained agent to current data to predict optimal settings for immediate impact.

python
1# Use the trained agent to predict control settings
2current_data = pd.read_csv("current_building_conditions.csv")
3optimal_settings = agent.predict(current_data)
4print(optimal_settings)

Automating Ad Campaigns with Reinforcement Learning

Discover how a mid-sized marketing agency achieved near-full automation and reduced media inefficiencies, powered by pi_optimal cutting-edge RL technology.

Cloud Platform
Scale RL to Production in Minutes

Effortlessly scale your RL solution from prototype to production in minutes. Train, deploy, monitor, and automate with ease.

Train

Streamline training with automated tools. Upload your data, and let the cloud find the best parameters.

Deploy

Deploy instantly with a ready-to-use API endpoint for seamless integration.

Monitor

Track your agent's performance in real-time with actionable insights.

Workflows

Automate retraining jobs to keep models up-to-date without manual effort.