Machine Learning Engineer, mechanical & non-mechanical domains

Our models run on the customer's own hardware, often with no internet connection. The customers are industrial groups and OEMs in India, Ireland, Germany and the USA.

A model that reads a machine is only useful if a technician can rely on it. So before we train anything, we write down how the machine works and how it fails. The model comes second.

What we build

Varahi is an engineering company in Pune. We build AI that runs on hardware the customer owns, inside their own building, and keeps working when the internet does not. This role sits in VarahiEdge, the line for companies that design, build and service machines:

What you will work on

• Predictive maintenance, anomaly detection and fault diagnosis on real machine data: vibration, temperature, current, CAN/OBD signals and service history.
• Evaluation sets built with domain experts, so we know a model works before a technician relies on it.
• Deploying small, optimised models to edge boxes on the shop floor.
• Optionally, 3D and CAD data for generative design and engineering tools.

What we are looking for

• 4–7 years of hands-on ML experience, with models that reached production.
• A mechanical, production, automobile or mechatronics background is strongly preferred.
• Strong Python, and PyTorch or TensorFlow.
• Time-series and signal processing.
• An understanding of how mechanical systems actually fail.

If your background is not mechanical, apply anyway — and tell us how you learned the machine you modelled.