Machine Learning Engineer – Document Intelligence & Applied GenAI

Описание

The landscape of AI is evolving rapidly, and PandaDoc is investing heavily in machine learning to power the next generation of intelligent document workflows. Our goal is to build scalable, production-grade AI systems that automate document understanding, extract structured data at scale, and enable new AI-first product experiences for tens of thousands of businesses.

As an ML Engineer focused on Document Intelligence and GenAI, you will design, train, evaluate, and optimize models that transform unstructured documents into high-quality structured data. You’ll work across the full stack of model development—datasets, training, inference, deployment pipelines—and help bring cutting-edge research into real production systems at scale.

What makes this role unique?

Document Intelligence at Scale: Your work will directly power PandaDoc’s core AI capabilities—from layout detection and OCR to structured extraction, retrieval, and document-based reasoning.High Ownership, High Impact: You will design end-to-end ML systems, influence roadmap decisions, and work closely with product, engineering, and design to define requirements and ship production AI features.Real-World ML Challenges: You’ll tackle model robustness, evaluation, latency, observability, RAG quality, model routing, and the complexities of deploying AI systems that must perform reliably on millions of documents.Deep GenAI Integration: You’ll experiment with frontier and open-source models, integrate vision–language systems, and build efficient pipelines for inference, guardrails, fine-tuning, and document-aware reasoning.

In this role, you will:

Model Development & EvaluationBuild and maintain evaluation frameworks for document models, LLMs, OCR, and structured extraction.Define metrics, benchmarks, and validation strategies for real-world document workloads.Dataset & Pipeline CreationDesign and curate high-quality datasets for supervised training, fine-tuning, and validation.Create scalable preprocessing pipelines for PDFs, scans, images, forms, and semi-structured documents.Model Training & Fine-TuningTrain and fine-tune transformer-based OCR, VLMs, layout models, and open-source LLMs for document understanding tasks.Optimize models for reliability, accuracy, and cost efficiency in production environments.Inference & DeploymentDeploy ML models with modern inference runtimes (vLLM, TGI, TensorRT, ONNX Runtime).Build guardrails, monitoring, and fallback mechanisms to ensure safe and predictable model behavior.RAG & Document ReasoningDevelop retrieval and chunking strategies tailored to document structures (tables, forms, multi-page PDFs).Optimize end-to-end RAG pipelines for semantic search, Q&A, and workflow automation.Cross-Functional CollaborationPartner with PMs, backend engineers, and product designers to define AI opportunities and translate requirements into technical solutions.

About you:

We are expanding our AI/ML function with an ML Engineer who specializes in document intelligence, vision–language models, and LLM-based extraction and reasoning. You should be comfortable with both traditional document AI approaches and cutting-edge GenAI workflows. You thrive in fast-moving environments, are self-directed, and enjoy solving practical ML problems that directly impact customers.

We’re looking for someone with experience in:

Vision transformers, layout models, and OCR systemsStructured extraction from complex documentsRAG for document-heavy workloadsOptimizing LLM pipelines for cost, accuracy, and throughputDeploying and benchmarking models in real production systems

Required Experience

5+ years of Python experienceExperience training, fine-tuning, and deploying traditional computer vision models for document intelligence tasks (layout detection, table extraction, OCR, information extraction)Hands-on experience with document understanding frameworks and models: Traditional document AI models (LayoutLM, Donut, DocFormer)Modern vision-language models with OCR capabilities (DeepSeek-OCR, LightOnOCR-1B, etc.)Experience deploying and optimizing models using inference frameworks such as vLLM (preferred), TGI, TensorRT, or ONNX RuntimeExperience applying LLMs to document intelligence workflows, including both frontier models and open-source alternativesStrong understanding of coordinate systems and spatial reasoning for absolute positioning and field detection in forms/documents

It would be awesome if you had:

Familiarity with PDF parsing libraries and document preprocessing pipelinesExperience fine-tuning open-source models for domain-specific document tasksKnowledge of evaluation metrics for document understanding tasks (F1, exact match, etc.)

Company Overview: 

PandaDoc empowers more than 67,000 growing organizations to thrive by taking the work out of document workflow. PandaDoc provides an all-in-one document workflow automation platform that helps fast scaling teams accelerate the ability to create, manage, and sign digital documents including proposals, quotes, contracts, and more.