Solutions

Unique Challenges

mapping-semiconductor

Mapping the Semiconductor Genome™

The Semiconductor Industry presents unique challenges for traditional machine learning and data mining approaches. Some unique characteristics of semiconductor processes are:

  • nonlinearity in most batch processes
  • multimodal batch trajectories due to product mix
  • process drift and shift
  • process steps with variable durations (often deliberately adjusted)
  • very complicated processes and process relationships
At StreamMosaic, we fully understand these unique challenges and have crafted solutions to tackle them.

We have unique solutions tailored to specific challenges facing the semiconductor industry:

  • Yield improvement for finFet’s
  • Yield improvement for 3-D NAND
  • Yield improvement through multi-patterning control
  • Yield improvement through overlay error compensation
  • Virtual Metrology
  • Fault Prediction and Classification
  • Predictive yield
  • Yield management
  • Run-to-run control
  • Wafer-to-wafer control
  • Real-time and in-situ control
  • Factory-wide control
  • Predictive Maintenance
  • Equipment excursion prediction
StreamMosaic emphasizes fully understanding the business problem before crafting a solution and plan.

The service and solution mix will be tailored to each individual client situation.
We have the semiconductor manufacturing expertise to understand how to organize the datasets and which models are most likely to be successful.

Our Methodology

our_approach

Focusing on understanding and solving the Business Problem

Our Customer Engagement Methodology Focuses on understanding and solving the Business Problem, then building the right Architecture around it.

Understanding the Specific Problem
We work closely with your team to fully understand the objectives, assess the current situation, determine the goals, and generate a plan.
Data Exploration & Understanding
We collect initial data and determining what data is available data, explore data, and determine data quality.
Data Preparation
We select and clean relevant data attributes and construct new attributes. Integrate data in preparation for modeling.
Modeling
We select modeling techniques, build and assess models.
Evaluation
We evaluate results and review next steps, likely to repeat some or all of the above steps (an iterative process).
Solution Deployment
We plan and implement deployment and business rules