Lumaril connects to the control systems already running in factories and other heavy plants, learns how each facility performs at its best, and continuously adjusts production for higher throughput, better yield, and less waste. It is aimed at capital-intensive operations in manufacturing, chemicals, biologics, oil and gas, agriculture, and polymers, where rigid controls and retiring operators leave output short of what the equipment can do. The launch matters now because that operator knowledge was largely never written down, and because Lumaril's claim is that existing US plants can add capacity faster than new factories alone. The software plugs into installed controls, adds commodity audio and visual sensors, and builds a live model of the facility. Classical control and machine learning choose the next move, and the platform changes setpoints, control logic, and coordination through a safety layer, with no rip-and-replace of hardware. Lumaril is paid on the production it adds. The launch film puts that work in operating plants. At The Arthur Companies, a family grain business, elevator operators describe catching problems by sound and sight after years on the floor. A family-owned plastics recycler in Portland describes the same constraint: the most experienced operators have decades in the plant, and handing that judgment to a new hire has taken years. Lumaril's team works on site with those operators so the model can learn why a decision was made. Felicis, an investor since the seed round, says the software is already live at multiple facilities. Co-founders Mark Elliot and Ben Duffield created Palantir Foundry. Elliot was commercial CTO at Palantir and later CTO at Theorem, and Duffield led the Foundry group. The company is coming out of stealth with a $66 million Series A at a $400 million valuation, led by Elad Gil, with Felicis, Lux Capital, Koch Disruptive Technologies, Pax Ventures, and General Catalyst.