Ertas for Document Classification

The Challenge

Organizations process thousands of documents daily — contracts, invoices, correspondence, reports, applications, and compliance filings — and routing each document to the right team or workflow depends on accurate classification. Manual classification is slow, inconsistent, and scales poorly. When a single misrouted document can delay a legal filing or lose a time-sensitive business opportunity, the cost of errors is significant.

Generic AI models struggle with document classification in specialized domains because they lack context about an organization's specific document taxonomy. A general model might distinguish between an invoice and a contract, but it cannot reliably differentiate between a master services agreement and a statement of work, or between a regulatory filing and an internal compliance memo. These fine-grained distinctions require domain knowledge that can only come from training on the organization's actual document corpus — exactly the kind of task that fine-tuning is designed to solve.

The Solution

Ertas enables organizations to fine-tune classification models on their own document taxonomy using real examples from their archives. With Ertas Studio, teams upload labeled document samples in JSONL format — where each entry maps document text to its correct category — and train a lightweight LoRA adapter that teaches the model to recognize the specific patterns, vocabulary, and structural cues that distinguish each document type in their taxonomy.

The fine-tuned model can be deployed as a classification endpoint through Ollama, vLLM, or Ertas Cloud, processing incoming documents in real time with sub-second latency. Because the model runs on your infrastructure, sensitive document content never leaves your network. Ertas Vault ensures that all training data and model artifacts are encrypted and access-controlled, meeting the data governance requirements of regulated industries. As the document taxonomy evolves — new categories are added, existing ones are split or merged — teams can retrain the model in Ertas Studio with updated examples and redeploy without any application changes.

Key Features

Example Workflow

A large insurance company receives 10,000+ documents daily across email, fax, and web portal channels. The documents include new claims, policy amendments, medical records, adjuster reports, and legal correspondence — each requiring routing to a different department. The team exports 50,000 labeled document examples from their archive and uploads them to Ertas Vault. In Ertas Studio, they fine-tune a 7B model with a LoRA adapter targeting their 28-category taxonomy. After training, the model achieves 96% classification accuracy on a held-out test set — compared to 71% from a generic model. The classifier is deployed as an API endpoint behind their document intake system, automatically routing each incoming document to the correct department queue with a confidence score. Documents below the confidence threshold are flagged for human review, creating a feedback loop that generates additional training data for future model improvements.