View ML engineer tasks

The machine learning engineer (ML engineer) is one of the most technically skilled roles. The ML engineer background should include completed end-to-end Java-based projects and experience in creating business applications, rolling them out to production, and dealing with environment challenges.

What is expected from ML engineer

The ML Engineer's tasks are as follows:

• Enforce a qualitative training set as a starting point for an ML solution:
  • Perform and explain Data Analyst (DA) key functions, such as document labeling and analyzing P/R/TP/FP/FN.
  • Identify problems with a training set and solve them by manipulating document sets.
• Improve models based on the data analysis in line with the WorkFusion workflow:
  • Analyze training results and understand what needs to be improved based on data analysis.
  • Know parameters that influence the overall model extraction quality and methods to improve it.
  • Know how to read extraction per field output statistics and what needs to be improved based on this analysis.
  • Understand typical AutoML pipeline: OCR > Labeling > Model training > AutoML SDK.
  • Possess practical skills in Search Engine settings for training (fields, experiments, time limit, max iterations).
  • Understand how to launch fixed model training or an improved training set for a previously trained model without Search Engine.
  • Be able to split a dataset and a test model on a test dataset.
  • Have business case implementation experience, with a delivered proof of concept as a final certification task as part of the the Automation Academy learning path.
• Use WorkFusion AutoML components as an ML solution:
  • Train models in Control Tower, set up Model steps and create Bot Tasks, and design sequential launches of training processes.
  • Deploy a model to make it available via Control Tower.
  • Transfer an ML solution through DEV > QA > UAT > PROD.
  • Understand the difference between the development and production environments and their usage strategy.
  • Solve issues in failed training processes.
  • Configure Mesos/Marathon and ZooNavigator settings.
  • Tune OCR capabilities, distinguish OCR errors, and fix them using AutoML SDK.
  • Know and use environment health checks.
  • Apply investigation skills when training fails.
• Apply AutoML SDK when needed.
• Understand the theoretical meaning of Annotators and Feature Extractors and how they relate to model fields.
• Be able to describe the AutoML SDK flow logically.

Role in RACI matrix

RACI stands for:

• R: Responsible
• C: Consult
• A: Accountable
• I: Informed

For ML Engineer responsibilities according to the RACI matrix, refer to the table below:

Area	Responsibility
Set up OCR environment and Business Processes	I
Create a Qualification Task for a customer (SME)	I
Prepare a training set	AC (technical help to DA if required)
Analyze and improve a training set	AC
Create a training set from customer data	R (with DA)
Set up ML environment	RA
Train a model	RA
Evaluate ML results and statistics	R (with DA)
Analyze metrics and provide post-processing description	CR (with DA)
Implement post-processing	RA
Generate an automation report for a customer	C