Shepherd: Reinforcement Learning Agent as a Service

Shepherd is an open-source web application allowing non-experts in Reinforcement Learning (RL) to easily use state-of-the-art RL techniques. It acts as a bridge between web clients, that connect to it over the network (using JSON commands sent over HTTP), and Reinforcement Learning agents running on our machines. A detailed technical documentation is available.

No specific dependencies must be installed on the client’s side, nor must a specific programming language be used. Our team regularly updates the algorithms used, following the advancements presented in the literature.

If you wish to know whether Reinforcement Learning techniques could be a good fit to your problem, or increase the performance of your process, our team proposes:

  • A first 2 hours meeting with our experts to identify your problem and needs, free of charge.
  • A 2 days feasibility study (go, no-go), at the end of which you will know if a solution, or a gain in performance can be achieved using RL.
  • A 1 to 2 weeks mission, at the end of which we will deliver a Reinforcement Learning solution tailored to your specific needs.

Feel free to contact us.