Senior Machine Learning Engineer, Synthetic Data & Document Understanding

📋 Job Overview

Join ABBYY and be part of a team that celebrates your unique work style. With flexible work options, a supportive team, and rewards that reflect your value, you can focus on what matters most – driving your growth, while fueling ours.

Our commitment to respect, transparency, and simplicity means you can trust us to always choose to do the right thing.

As a trusted partner for purpose-built AI and intelligent automation, we solve highly complex problems for our enterprise customers and put their information to work to transform the way they do business. Over 10,000 customers trust ABBYY, including many Fortune 500 ones. You will work on further developing a portfolio already containing client names such as DHL, Johnson & Johnson, FDA, DMV, PwC, KeyBank, Spotify, and H&R BLOCK.

We are seeking a Senior Machine Learning Engineer – Synthetic Data & Document Understanding to own the synthetic data generation track within ABBYY’s Document AI Data team.

This role focuses on building generative pipelines that produce high-quality, diverse, and realistic synthetic training data at scale. You will ensure synthetic data meaningfully improves downstream model performance by maintaining strong alignment with real-world document structures, formats, and statistical properties.

This is an ideal role for engineers who combine deep generative modeling expertise with rigorous data quality evaluation and production engineering skills.

🏢 The Role

Key Responsibilities

• Design and implement pipelines that analyze real documents to inform high-fidelity synthetic data generation
• Build generative systems capable of producing documents across diverse formats, layouts, and domains
• Develop evaluation frameworks to ensure synthetic data maintains distributional fidelity and diversity
• Research and apply generative modeling techniques suited for document AI training
Identify and mitigate quality issues to ensure synthetic data is effective for downstream model training
• Partner with Modeling teams to measure the impact of synthetic data on model performance
• Own the synthetic data generation track end-to-end, from architecture to quality validation
• Drive architectural decisions balancing quality, diversity, scale, and cost efficiency
• Define and maintain data quality metrics and generation dashboards
• Collaborate with annotation teams to ensure compatibility with downstream pipelines
• Contribute to roadmap planning alongside Principal-level leadership
• Build scalable pipelines capable of generating millions of synthetic training examples
• Implement post-processing, filtering, and validation mechanisms to remove low-quality outputs
• Design cost-efficient workflows balancing compute, quality, and throughput
• Develop monitoring systems to detect distribution shifts or quality degradation
• Collaborate with Platform teams on compute orchestration, storage, and scheduling

• 🔧 Key Skills

• Deep expertise in Vision-Language Models and document understanding (layout, structure, semantics)
• Strong knowledge of generative modeling for structured and semi-structured data
• Understanding of what makes synthetic data valuable: distributional fidelity, diversity, realistic noise patterns, domain coverage
• Strong programming skills in Python with experience in PyTorch or similar frameworks
• Experience evaluating data quality via automated metrics and downstream impact
• Familiarity with large-scale data pipelines, cloud environments, and experiment tracking

• 🎁 Benefits

  Comprehensive medical, accidental, and life insurance

  Weekly wellness sessions to support your physical and mental well-being

  A generous paid time off policy

  ABBYY is an Equal Employment Opportunity employer valuing diversity in the workplace. Learn more about our commitment to Diversity and Inclusion in the careers section on our website.

• 🎯 The Team

  ABBYY is a company with over 35 years of experience in the technology market; Over 10,000 customers including Fortune 500 ones with names such as DHL, Johnson & Johnson, FDA, DMV, PwC, KeyBank, Spotify, and H&R BLOCK. We have specialized in document capture by being the first low-code document capture platform. Our Machine Learning, Natural Language Processing, Computer Vision, and AI technologies can transform any document into any format.

• ⭐ Nice to Have

• 📧 Visa & Immigration