Led a large-scale visual data production and quality assurance project involving approximately 110,000 collected images and 100,000 delivered image items.

The work covered image collection and screening, structured reconstruction of tables and flowcharts, question-answer data preparation, quality inspection, rework management, packaging, and final delivery.

I was responsible for the overall project execution, workflow organization, automation tooling, quality control, and delivery coordination.

A team of approximately five people participated in quality inspection activities.

The project required a production process that could handle large volumes of data while maintaining traceability, consistency, and reliable delivery.

Processing a large visual dataset is not simply a matter of collecting files.

Different content types require different handling, while errors introduced during processing, annotation, or quality inspection can affect downstream data usability.

The project presented several operational and engineering challenges:

Processing and organizing approximately 110,000 collected images

Reconstructing structured information from tables and flowcharts

Preparing question-answer data associated with visual content

Maintaining consistency across multiple production batches

Identifying incorrect or incomplete AI-assisted outputs

Coordinating human quality inspection and rework

Managing annotations, manifests, and delivery packages

Verifying package integrity before final delivery

The main engineering objective was to establish a repeatable and verifiable workflow rather than relying entirely on manual processing.

Managed the overall workflow from image acquisition to final delivery.

Organized image collection, screening, structured content preparation, quality inspection, rework, and data packaging.

Coordinated production activities and quality inspection responsibilities across the participating team.

The process ultimately covered approximately 110,000 collected images, with approximately 100,000 delivered.

Developed automation tools using Python and PowerShell to support large-scale data processing operations.

The tooling covered:

Image preprocessing and screening

Batch file splitting and reassembly

Annotation and manifest processing

Structured field modifications

Batch image operations

Quality inspection interfaces

Rework and correction utilities

Delivery package reconstruction

Pre-delivery validation

These tools reduced repetitive manual operations and supported more consistent processing across data batches.

Organized the processing of visual content containing tables, flowcharts, and other structured information.

The workflow included reconstructing structured content and preparing question-answer data associated with the source material.

Automation was used where appropriate, with human review retained for tasks requiring interpretation and correctness checks.

Introduced large language models to assist with question-answer preparation and quality inspection.

Recognizing that model-generated outputs could contain interpretation errors, established additional verification steps rather than treating AI output as automatically correct.

The quality control process included:

Cross-checking outputs using two models

Human quality inspection

Error correction and rework

Independent blind inspection

Additional verification after rework

This approach combined automated assistance with human judgment and independent review.

Developed and used tools for handling annotations, manifests, batch restructuring, and data package validation.

Organized quality feedback and correction workflows so problematic items could be reworked and checked again before delivery.

The final delivery process included package reconstruction and validation to reduce missing files, inconsistent fields, and packaging errors.

Completed a large-scale visual data production project involving approximately 110,000 collected images and 100,000 delivered image items.

Established an operational workflow combining Python/PowerShell automation, AI-assisted processing, human quality inspection, independent verification, and structured delivery.

The project demonstrates experience in:

Large-volume file and image processing

Production workflow design

Data preparation and structured content handling

Automation tool development

AI-assisted quality assurance

Human review and rework coordination

Batch packaging and delivery validation

Multi-person production management

The key engineering contribution was building and operating a repeatable data production process with multiple quality control stages, rather than treating data preparation as a collection of isolated manual tasks.

Python · PowerShell · Batch Processing · Image Processing · Annotations · Manifests · AI-Assisted QA · Human Quality Control · Data Validation · Delivery Automation