While the chemical industry is facing a disruptive transition into renewable raw materials and sustainability, the design and engineering of new (bio)chemical processes is still based on previous processes, patents or bibliography. That makes new projects last for years and very expensive as it requires a group of chemical engineers to perform the design, optimization and analysis of such processes on an industrial scale, to estimate the economic and environmental feasibility.
Reinforcement learning is an emerging AI technology that is promising in complex optimization and sequential decision making, specially in fields or scenarios that can be simulated, like videogames for example, in where the AI agent learns to interact and take decisions to optimize the performance, just by knowing the rules of the game and playing it a lot of times.
This opens a new world of possibilities in science discovery and engineering in which we know the first-principles (often described by mathematical models) so the behaviour of processes can be simulated in a computer, and some performance metrics can be numerically quantified, which for a chemical process could be the economic costs and sustainable metrics for products and processes like the life cycle analysis.
Intemic is applying that concept to chemical process design, where the agent learns how to design economic and sustainable processes to transform renewable materials like biomass or waste into valuable products like biofuels, bioplastics or bioproducts for the pharmaceutical and cosmetics industry, by performing millions of simulations and learning from them.
We think this technology can accelerate our transition into a circular and sustainable bioeconomy, helping to mitigate emissions and climate change.