For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

Invite to Job
M
Mt

Mt

Agency

Principal Technical Consultant (MT Digital) — Multilingual Document Processing & SFT Alignment Pipeline

Australia flagBrisbane, Australia

Key Skills

Software

CVATCVAT
LabelboxLabelbox
Internal/Proprietary Tooling
Other

Top Subject Matter

Multilingual document AI
SFT dataset construction and alignment
High-density document layout annotation and OCR/Document-AI alignment

Top Data Types

ImageImage
VideoVideo
TextText
DocumentDocument

Top Task Types

Entity (NER) ClassificationEntity (NER) Classification
Prompt + Response Writing (SFT)Prompt + Response Writing (SFT)
Computer Programming/CodingComputer Programming/Coding
Data CollectionData Collection
Text GenerationText Generation

Company Overview

MT Digital is a premium data operations consultancy based in Brisbane, Australia. We specialize in providing highly secure, 100% onshore data labeling and AI training architecture for organizations handling sensitive intellectual property or strictly regulated information.

Security

Security Overview

At MT Digital, data sovereignty and absolute privacy are the core pillars of our technical infrastructure. Unlike standard annotation providers reliant on distributed public APIs and offshore workforces, our operations are 100% onshore in Brisbane, Australia, and executed within tightly controlled environments. Our core security controls include: 1. Sovereign Air-Gapped / Offline Operations: We architect local, self-hosted AI inference nodes and orchestration pipelines running entirely behind dedicated hardware firewalls. Sensitive client datasets, legal documentation, and financial files are processed locally on-premise, ensuring zero data leakage to public third-party LLM clouds. 2. Access Control & Network Hardening: Inter-machine communication across our operational cluster is strictly containerized and managed via encrypted Secure Shell (SSH) networking. Internal firewalls enforce strict IP whitelisting to guarantee that port exposure is limited entirely to validated internal nodes. 3. Secure Infrastructure Management: Operating under the principles of the Australian Cyber Security Centre's (ACSC) Essential Eight, we enforce restricted administrative privileges across all endpoints, employ multi-factor authentication (MFA) for platform access, and implement isolated local sandbox environments for all code execution and function-calling operations. By eliminating the third-party dependencies of public cloud routing, we ensure that our clients’ proprietary intellectual property and regulated corporate data remain fully localized, audited, and structurally secure throughout the training and evaluation lifecycle.

Security Credentials

ISO 27001

Labeling Experience

Principal Technical Consultant (MT Digital) — High-Density Document AI Layout Analysis & OCR Alignment

DocumentDocument

Executed a high-throughput visual annotation effort to align Document-AI/OCR outputs with page geometry and semantic strings for dense financial documents. Optimized context processing by segmenting low-resolution archival text, multi-column forms, and structured page elements offline. Achieved >95% character accuracy thresholds while maintaining complete offline operation and data sovereignty. • Segmented dense financial documentation and multi-column forms for layout understanding. • Mapped page geometry to semantic strings to improve Document-AI alignment. • Processed low-resolution archival text with geometry-aware annotation. • Delivered high-accuracy (>95%) offline character thresholds for training/evaluation readiness.

2024 - Present

Principal Technical Consultant (MT Digital) — Multilingual Document Processing & SFT Alignment Pipeline

OtherDocumentDocumentPrompt + Response Writing (SFT)Prompt + Response Writing (SFT)

Engineered an SFT alignment pipeline that translated and converted complex multilingual scanned assets into contextually precise English Markdown. Built validation workflows to evaluate supervised fine-tuning readiness using local vision-language processing. Designed offline-safe dataset preparation and prompt alignment suitable for training and evaluation in sovereign on-prem environments. • Translation and Supervised Fine-Tuning validation pipeline for non-English structural documents. • Processed scanned documents via local vision-language frameworks to clean and map content to English Markdown. • Used reasoning nodes to preserve contextual structure during transformation. • Focused on high-accuracy results for complex, non-English document formats.

2024 - Present
Labelbox

Mass Image Training for Document Layout & Text Accuracy

LabelboxLabelboxImageImageObject DetectionObject DetectionText GenerationText Generation

Executed a high-throughput visual annotation initiative focused on optimizing Document-AI and Vision-Language Model (VLM) context processing. The scope included segmenting high-density financial documentation, multi-column forms, stamped corporate paper trails, and low-resolution archival text to train optical layout compression layers. Performed granular structural mapping to link physical page geometry with extracted semantic strings. Strict verification guidelines were implemented across a 10,000+ item dataset, utilizing programmatic validation layers to eliminate character distortion errors, optimize spatial coordinates for bounding logic, and maximize downstream OCR reading accuracy across variant document orientations.

2026 - 2026

Enterprise Data Operations Lead (Contract) (Geospatial Resource Project) — Secure Geospatial Data Pipeline

TextTextData CollectionData Collection

Built an automated geospatial data ingestion pipeline that transformed raw unstructured drilling logs and core sample records into structured schema-validated JSON for local testing. Implemented processing suitable for downstream labeling and analysis workflows under strict data residency constraints. Ensured searchable structured outputs without public cloud routing across the contract period. • Converted raw geospatial text records into structured JSON arrays. • Applied schema validation to prepare data for structured labeling/analysis. • Maintained strict local data residency compliance. • Optimized search accuracy for an 8-month window using local processing.

2024 - 2024

Geological Viability Data Architecture for Mining Operations

Internal/Proprietary ToolingTextTextData CollectionData Collection

Synthesized, parsed, and cataloged extensive archives of unstructured drilling logs, core sample reports, and historical geospatial assays to build a structured training dataset for localized resource modeling. Extracted entity metrics including lithological compositions, grade parameters, and depth matrices via custom automated terminal scripts and private ingestion pipelines. Structured the data into unified JSON arrays to enable seamless local downstream model testing. Compliance & Certification Note: This technical consulting project operated strictly as an IT system engineering, data infrastructure, and programmatic annotation workflow; it did not require, nor constitute, formal JORC resource certifications or professional geological signing authority.

2024 - 2024