<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6" -->

---
title: Your Agreements Are a Database You Can't Query — Hiral...
description: Docusign and NVIDIA presented their partnership on extracting structured data from millions of unstructured agreement documents. Docusign processes about a...
canonical: https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Your Agreements Are a Database You Can't Query — Hiral Shah, Docusign & Sean Sodha, NVIDIA | daily.dev
og:description: Docusign and NVIDIA presented their partnership on extracting structured data from millions of unstructured agreement documents. Docusign processes about a...
og:url: https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6
og:image: https://api.daily.dev/og/posts/3dnkXvwu6.png
og:image:alt: Your Agreements Are a Database You Can't Query — Hiral Shah, Docusign & Sean Sodha, NVIDIA
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Your Agreements Are a Database You Can't Query — Hiral Shah, Docusign & Sean Sodha, NVIDIA

**[AI Engineer](https://daily.dev/sources/aidotengineer)** · 18 min read · 0 upvotes · 0 comments

## Summary

Docusign and NVIDIA presented their partnership on extracting structured data from millions of unstructured agreement documents. Docusign processes about a million agreements daily and cites a Deloitte study estimating $2 trillion in agreement value going uncaptured due to manual review. NVIDIA's Nemotron Parse model, an 850-900 million parameter vision-language model, was built specifically for table extraction (pricing tiers, SKUs, SLAs, rate cards) where generic VLMs and traditional OCR tools fail on merged cells and complex layouts. The model is a lightweight extractor (not a generator), deployable via NVIDIA NIM, and reportedly extracts tables 20x faster than other open-source models tested. A live demo showed Docusign's Agreement Manager uploading a contract, auto-extracting structured pricing tables, and exporting to CSV/API. The teams discussed running FP16 now with plans to move to FP8/NVFP4 on Blackwell hardware, and outlined next steps including a Nemotron Retriever integration and the NVIDIA agent toolkit for scaling into production agents.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://www.youtube.com/watch?v=_gvamfT8H-w>

## Questions this post answers

### What is the NVIDIA Nemotron Parse model and what is it used for?

Nemotron Parse is an 850-900 million parameter vision-language model built by NVIDIA specifically for document layout and table extraction rather than text generation. It takes a document image as input and outputs semantic formatting, reading order, and preserved table structure, addressing failures in generic VLMs and OCR tools on merged cells and nested tables. It can be served via NVIDIA NIM or as an LLM, and Docusign uses it inside Agreement Manager to extract pricing tables, SKUs, and SLAs from contracts.

_Teams evaluating document extraction models can track new releases like this on daily.dev._

### Why do generic vision-language models fail at extracting tables from contracts?

Generic VLMs and traditional OCR tools read text line by line, which breaks the structural concept of a table, especially with merged cells, nested tables, or unboundaried columns and rows. This causes significant errors when extracting vital contract terms like pricing tiers, SKUs, and SLAs, forcing legal, procurement, and sales teams to spend hours manually locating data that a purpose-built table extraction model can pull out automatically.

_Engineers comparing document AI approaches can follow model benchmarks like this via daily.dev._

### How much faster is NVIDIA's Nemotron Parse model at table extraction compared to other open-source models?

Nemotron Parse extracts tables about 20 times faster than other open-source models tested, measured by tables extracted per second. This speed advantage, combined with its smaller size of roughly 850-900 million parameters, gives lower latency and lower cost, which matters at Docusign's scale of processing around a million agreements per day.

_Developers benchmarking extraction models for scale can keep tabs on results like these through daily.dev._

## Similar posts on daily.dev

- [Nemotron Labs: How AI Agents Are Turning Documents Into Real-Time Business Intelligence](https://daily.dev/posts/nemotron-labs-how-ai-agents-are-turning-documents-into-real-time-business-intelligence-zanlt6vkk) · NVIDIA · 1 upvotes · 0 comments
- [How to Build a Document Processing Pipeline for RAG with Nemotron](https://daily.dev/posts/how-to-build-a-document-processing-pipeline-for-rag-with-nemotron-1v2hotr8y) · NVIDIA Developer · 1 upvotes · 0 comments
- [Why Your Agents Can’t Read Enterprise Documents — and How to Fix It](https://daily.dev/posts/why-your-agents-can-t-read-enterprise-documents-and-how-to-fix-it-lexemrbet) · databricks · 0 upvotes · 0 comments
- [Beyond Semantic Similarity: Introducing NVIDIA NeMo Retriever’s Generalizable Agentic Retrieval Pipeline](https://daily.dev/posts/beyond-semantic-similarity-introducing-nvidia-nemo-retriever-s-generalizable-agentic-retrieval-pipe-3dknsiul1) · Hugging Face · 2 upvotes · 0 comments

---

Tags: [#nvidia](https://daily.dev/tags/nvidia), [#rag](https://daily.dev/tags/rag), [#vlm](https://daily.dev/tags/vlm)

[View this post on daily.dev](https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Your Agreements Are a Database You Can't Query — Hiral Shah, Docusign \u0026 Sean Sodha, NVIDIA","url":"https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6"},"datePublished":"2026-09-16T15:57:27.165Z","dateModified":"2026-09-16T15:57:53.442Z","description":"Docusign and NVIDIA presented their partnership on extracting structured data from millions of unstructured agreement documents. Docusign processes about a...","image":"https://i.ytimg.com/vi/_gvamfT8H-w/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/_gvamfT8H-w/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"nvidia,rag,vlm","timeRequired":"PT18M","video":{"@type":"VideoObject","name":"Your Agreements Are a Database You Can't Query — Hiral Shah, Docusign \u0026 Sean Sodha, NVIDIA","description":"Docusign and NVIDIA presented their partnership on extracting structured data from millions of unstructured agreement documents. Docusign processes about a...","thumbnailUrl":"https://i.ytimg.com/vi/_gvamfT8H-w/sddefault.jpg","uploadDate":"2026-09-16T15:57:27.165Z","duration":"PT18M","url":"https://api.daily.dev/r/3dnkXvwu6","embedUrl":"https://www.youtube.com/embed/_gvamfT8H-w"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Your Agreements Are a Database You Can't Query — Hiral Shah, Docusign \u0026 Sean Sodha, NVIDIA"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/your-agreements-are-a-database-you-can-t-query-hiral-shah-docusign-sean-sodha-nvidia-3dnkxvwu6#faq","mainEntity":[{"@type":"Question","name":"What is the NVIDIA Nemotron Parse model and what is it used for?","acceptedAnswer":{"@type":"Answer","text":"Nemotron Parse is an 850-900 million parameter vision-language model built by NVIDIA specifically for document layout and table extraction rather than text generation. It takes a document image as input and outputs semantic formatting, reading order, and preserved table structure, addressing failures in generic VLMs and OCR tools on merged cells and nested tables. It can be served via NVIDIA NIM or as an LLM, and Docusign uses it inside Agreement Manager to extract pricing tables, SKUs, and SLAs from contracts. Teams evaluating document extraction models can track new releases like this on daily.dev."}},{"@type":"Question","name":"Why do generic vision-language models fail at extracting tables from contracts?","acceptedAnswer":{"@type":"Answer","text":"Generic VLMs and traditional OCR tools read text line by line, which breaks the structural concept of a table, especially with merged cells, nested tables, or unboundaried columns and rows. This causes significant errors when extracting vital contract terms like pricing tiers, SKUs, and SLAs, forcing legal, procurement, and sales teams to spend hours manually locating data that a purpose-built table extraction model can pull out automatically. Engineers comparing document AI approaches can follow model benchmarks like this via daily.dev."}},{"@type":"Question","name":"How much faster is NVIDIA's Nemotron Parse model at table extraction compared to other open-source models?","acceptedAnswer":{"@type":"Answer","text":"Nemotron Parse extracts tables about 20 times faster than other open-source models tested, measured by tables extracted per second. This speed advantage, combined with its smaller size of roughly 850-900 million parameters, gives lower latency and lower cost, which matters at Docusign's scale of processing around a million agreements per day. Developers benchmarking extraction models for scale can keep tabs on results like these through daily.dev."}}]}
```

