GMA/SSA

GMA/SSA

Agentic AI for semiconductor process optimization

Agentic AI for semiconductor process optimization

Company

Aitomatic

Position

Head of Product & Design

Year

2024

Market

APAC & US, B2B

01

Problem

Tokyo Electron, one of the world's leading semiconductor equipment companies, partnered with Aitomatic to pilot a multi-agent product for semiconductor process optimization.

TEL engineers relied on customer specifications, historical runs, process data, and expert judgment to define etching parameters. The process was slow, difficult to standardize, and costly when decisions went wrong. A failed adjustment could delay production or waste a wafer batch, costing approximately $10K to $20K per cycle.

02

Solution

GMA/SSA helped engineers structure optimization work, evaluate parameter recommendations, and correct weak agent logic before applying a recommendation.

For each problem, a coordinating agent formed a team of relevant specialist agents. Each specialist was conditioned with a defined role, technical context, operating constraints, and success criteria. Engineers could inspect the reasoning, adjust conditions, identify missing or faulty logic, and correct the relevant agent without restarting the full workflow.

03

my role

As Head of Product & Design, I shaped the agent team model, conditioning framework, decision workflow, and interaction model for diagnosing and correcting agent behavior. I directed the design execution with engineering and validated the product with TEL engineers against real technical workflows and production constraints.

04

impact

In pilot validation, GMA/SSA reduced the time required to reach an approved parameter set by 30%. The pilot showed approximately $20K in potential savings per etching cycle and more than $10M in projected annual savings across customer sites.

Create GMA

Create GMA

Create GMA

Select SSA

Select SSA

Select SSA

Create Plan

Create Plan

Create Plan

Define Task

Define Task

Define Task

Choose Next Step

Choose Next Step

Choose Next Step

Set Condition

Set Condition

Set Condition

Filled Condition

Filled Condition

Filled Condition

Q&A with GMA/SSA

Q&A with GMA/SSA

Q&A with GMA/SSA

SSA Lacks Knowledge

SSA Lacks Knowledge

SSA Lacks Knowledge

Contribute Knowledge

Contribute Knowledge

Contribute Knowledge