Physical Intelligence for Manufacturing.

We connect the shop floor, trace every material, kilowatt, process step and quality result, and turn it into decisions your team can act on.

We start where measuring is hardest: metal casting. Running today in iron foundries in India and the United States.

What we trace

Every plant runs on four flows, and on the machines that carry them. We trace all of it, from the yard to the finished part. Here is how it looks in an iron foundry.

Materials: Every lot, every heat, every part.
Power: Energy per heat and per tonne.
Processes: Every step and every wait, timed.
Quality: Causes, actions and proof.
Equipment: How every machine wears, and when to act.

Materials: Where did this part's metal come from?

Cameras follow every crane lift and every charge car, so each heat carries its own recipe: which bays, which lots, what weight. When a casting is rejected, you trace it back to its source in seconds. Captured by: Cameras and gateways on the crane and charge cars. You get: Yield and losses by lot and by heat.

Power: What does a good tonne really cost in energy?

Meters and furnace signals give the energy of every heat. We split it into melting, holding and idle time, so you see where the power goes and what each tonne of good castings really costs. Captured by: Energy meters, furnace controls, gateways. You get: kWh per tonne, per heat and per shift.

Processes: Where does the time go between steps?

Every step and every wait is timed: waiting for pour, treatment, transfer, pouring. Slow starts and long waits show up the same day, with the cost of each delay. Captured by: Cameras and machine data. You get: Cycle times, bottlenecks and delays for every heat.

Quality: Why did rejects rise, and what should we change?

Each defect is linked to the conditions behind it. You get a recommended change, then a check that the change worked. Autonomous labs and smart instruments send test results in minutes, not after the shift. Captured by: Autonomous labs, smart instruments, inspection records. You get: Causes, actions and proof of effect.

Equipment: When will this machine need attention?

Every machine has its own physics, and it changes as the machine wears: a lining thins, a bearing tires, a coil loses efficiency. We keep a physics model of each one and watch how it drifts, so maintenance happens before the breakdown, not after it. Captured by: Energy meters, vibration and temperature sensors, gateways. You get: Remaining life and the right time to service every machine.

We build the devices, and the intelligence they feed.

Each device sends its own data from the floor, and your people's know-how goes in with it. The intelligence layer pools all of it into one record for every heat, lot and part, and sends approved changes back to the floor.

Gateways: Read PLCs, meters and machines, old or new, and send every signal securely from the floor.
PC agents: Run on the plant's own computers and pick up spectrometer results, SCADA logs and spreadsheets the moment they are saved.
Autonomous labs: Test sand and metal on their own and send each result with its sample and its time.
Inline sensors: Measure moisture, temperature, weight and energy inside the process, as it happens.
Cameras: Follow cranes, ladles and moulds, and turn what they see into timed events.
Plant knowledge: The know-how your people built over decades: how the plant really runs, what went wrong and what fixed it. We capture it, so it stays when people move on and the intelligence reasons with it.
The intelligence layer: Pools every stream and every lesson on one timeline, ties them to each heat, lot and part, and keeps a physics model of every machine. It finds the causes and explains them in plain words.
Back to the floor: Approved changes go back to the machines through the same gateways: a setpoint, a hold, a recipe adjustment. Your team stays in charge of every one.

What we build

Plant twins: One live model of the plant: the records it already keeps, the data our products add and the physics of every machine.
Gateways: Secure links to PLCs, machines and instruments, old or new. Nothing is ripped out.
Autonomous labs: Robotic labs that run routine tests on their own, the same way every time.
Sensing systems: Cameras and smart instruments that measure what manual records miss.
AI that explains: Plain answers with the evidence behind them. Your team decides what changes.
Built with equipment makers: We turn machines and instruments into connected products, together with the companies that make them.

How we work with manufacturers and equipment makers

Walk the floor. We map the process, its records and the know-how of the people who run it.
Start with what you have. Your existing logs and files go in first, and the first findings follow.
Add sensing where data is missing. Cameras, gateways, labs and instruments, installed and supported by us.
Move from advice to control. At your pace: advice first, then approval, then automatic action within limits you set.

We turn your machines into connected products with you, sold under your name or alongside ours. With Versatile Equipments, manual sand testers became connected instruments and an autonomous sand lab.

Proven for metal casting

Molten iron, sand, heat and dust make a foundry one of the hardest places to measure anything. Our first complete stack runs there today, in iron foundries in India and the United States. The same approach carries into any industry that transforms materials.

Foundry Twin: One live model of the plant, built from the records it already keeps.
Floor Master: Cameras on the floor that turn what they see into timed events.
Black Box: An autonomous sand lab that runs routine tests the same way every time.
MicroVision: Vision inspection that measures what manual records miss.
Metallographic Black Box: Sample prep and metallography run by robot, results in minutes.
Versatile Equipments: Manual sand testers rebuilt as connected instruments.

The team

Built by engineers and researchers coming from Harvard University, Indian Institute of Science, University of Southampton, Queen Mary University of London and Intel.

Talk to us

Write to info@sahaslabs.com or use the contact form.