Why virtual metrology is the right first agentic AI use case for a fab
Virtual metrology has the data, the feedback loop and the bounded blast radius that a first agentic AI deployment needs.
Semiconductor Manufacturing × Agentic AI
I'm Jnanadarshan (JD) Nayak. I head Semiconductor Strategy & the Semiconductor Manufacturing Practice at TCS. I design how AI, and increasingly agentic AI, gets infused into fab operations. Outside client work I build software, simulators and educational tools, contribute to open source, run a self-hosted homelab, and design and print my own hardware.
Focus areas
The next productivity step in manufacturing will not come from more dashboards. It will come from systems that perceive, reason and act inside the production environment, with engineers supervising rather than operating.
Virtual metrology, defect root-cause analysis, trace data FDC, chamber matching and MES-integrated automation, grounded in real process and equipment knowledge.
Making agent systems trustworthy, governable and repeatable enough to proliferate across a factory and an enterprise, with autonomy that is earned, not declared.
Simulators, dashboards, agents and data pipelines. Small, sharp tools that work in a browser or a terminal and can be understood in an afternoon.
Autonomy is staged. Each step has its own evidence, thresholds and approvals before the next one begins.
Advise
The agent investigates and explains. Engineers decide and act.
Recommend with approval
The agent proposes a specific action. A human approves it at a defined gate.
Act within bounds
The agent acts inside explicit limits, with confidence thresholds and rollback paths.
Act and report
The agent owns a bounded task end to end. Humans supervise exceptions and set policy.
Beyond client work
Most of what I advise on, I have first prototyped myself.
Fixes, documentation and small tools that solve a problem someone else will hit next.
A small production environment I am allowed to break: virtualisation, storage, segmented networking, security monitoring and home automation.
Parametric design in OpenSCAD, 3D printing, embedded devices and hardware hacking.
Notes on AI in the fab, agent architecture, open source tooling and the homelab. Every fix gets written down.
Virtual metrology has the data, the feedback loop and the bounded blast radius that a first agentic AI deployment needs.
An agent that reasons over a structured, versioned description of your process outperforms a larger model that is guessing.
Autonomy should be earned through evidence and approvals. Here is the four-stage path I design agents around.
Agentic AI will not stay a pilot. Here are the five stages from a single advise-only agent to supervised autonomous operations.
I'm happy to talk about fab automation, agentic AI architecture and MES modernisation.