Factories across the globe are losing between 3% and 20% of their revenue due to inefficient planning of their production processes. Phantasma Labs says it has developed software that helps manufacturing companies of all sizes efficiently plan and schedule their processes. Built on the foundations of Reinforcement Learning AI, the Berlin-based startup says its software generates production plans that are up to 30% better and created 100 times faster than traditional methods.
As manufacturing companies focus on becoming more efficient and competitive, the demand for advanced production planning tools has increased exponentially. The global market for production planning software is estimated to be between $12 billion and $15 billion.
The uniqueness of the Berlin-based startup’s approach lies in its independence from Big Data, making its technology applicable to not just large corporates but SMEs and mid-caps. Think of it as a route to speeding up digitalization without a two-five year waiting period, says the company.
“Our mission is to make factories more efficient,” says co-founder and CEO Ramakrisha Nanjundaiah. “A lot of potential is untapped because companies don’t have the right tools and the barriers for implementing AI are really high. We offer AI methodologies that allow AI to be implemented where it has not been actionable before because of cost constraints and data constraints.”
Instead of training AI on data Phantasma Labs trains them on simulations. Nanjundaiah uses the analogy of training AI to play chess. “You could collect data on a million different chess players moves to train the AI, which be really expensive and result in an average player, or you can build a simulation of the game, put in the rules and let the AI agent make good moves and bad moves and learn how to win the game,” he says. Phantasma does the same with factories. It simulates all the things that can happen on a factory floor such as key staff calling in sick, a machine breaking down or if there is suddenly higher demand on a particular line. The AI learns from that and helps clients better anticipate change, says Nanjundaiah.
Since its technology does not require clean, organized Big Data Phantasma says it can develop a specific use case for clients within about four weeks. Deployment can be either on premise or in the Cloud.
Customers include automotive tier one and tier two suppliers, large OEMs and machine tool manufacturers. The company just raised a seven-figure investment round of financing It is already active in the DACH market and will use the new funds to expand into the U.S. and Nordic markets.
Competitors include established German market players Dualis and MPDV as well as newer players such as Belgium’s MangoGem.
Phantasma Labs, which was launched in 2020 with a focus on the automotive sector, pivoted in 2023 to focus on the industrial sector. Nanjundaiah says that along with the use of Reinforcement Learning AI, one of the team’s big differentiators is that it has both deep expertise in AI and Machine Learning as well as domain expertise. His own background is in computing with large scale simulations that involve complex physics and numerics. He was previously involved in an innovative project in the Middle East focused on designing one of the world’s largest structures. Other team members, which come from Bosch, Oracle and the Technical University of Munich, have expertise in building AI models for production.
