Autonomi AI

Today, we're excited to unveil ??????-?? - a highly-specialized and versatile visual understanding API that excels at extracting structured JSON from images, videos and documents. Unlike existing visual APIs that are trained to chat with images,?VLM-1 is hyper-specialized to perform structured prediction for virtually any visual domain it's trained on (e.g. financial docs/presentations, watching tv, web-automation, sports/media analytics). And unlike most solutions on the market today, you can fine-tune and deploy a new VLM-1 model customized for your domain in just under an hour on our optimized platform. Some key features of VLM-1 we’re particularly excited about: 1. ?? Structured Outputs: VLM-1 provides structured predictions (e.g. JSON) for your images/videos/documents, allowing you to easily automate visual tasks with strongly-typed and validated outputs. 2. ?? Hyper-specialized: VLM-1 can be fine-tuned for specific visual domains, allowing you to achieve the desired accuracy for your use-case with enterprise-level SLAs. 3. ?? Scalable: VLM-1 is optimized to be cost-effective for high data volumes, enabling you to scale your visual automation workflows without being rate-limted or incurring large bills. 4. ??? Private: VLM-1 can be deployed on-prem or in a private cloud, allowing you to keep your data private and secure, and work with privacy-sensitive material. If you're interested in using VLM-1 (??????????-??????????????) for your structured visual understanding workflows, sign up for early API access and check out a few of our cookbook examples (see below). We're currently working with a select group of partners to fine-tune the API and would love to hear from you if you have a specific high-volume use-case in mind. ?? Demo: https://lnkd.in/guN--aEd ?? Sign-up for API access: https://lnkd.in/grmJi8Cu ???? Cookbook: https://lnkd.in/g75fHdRC We fundamentally believe this is a new paradigm-shift in visual learning and reasoning, and we'll be showcasing a few of VLM-1's capabilities for both the cloud and the edge over the coming weeks. More demo links in the thread below ?? #foundationmodel #computervision #pydantic #ai #genai

???Newly re-designed VLM Run website just dropped! https://vlm.run ? If you’re looking to leverage state-of-the-art Vision-Language Models (VLMs) or simply up-level your Visual AI stack, drop us a line. We’ve got some exciting updates coming soon. #computervision #vlm

???Newly redesigned VLM Run website just dropped! https://vlm.run ? If you’re looking to leverage state-of-the-art Vision-Language Models (VLMs) or simply up-level your Visual AI stack, drop us a line. We’ve got some exciting updates coming soon. #vlm #visualai #computervision

?? VLM Run is excited to officially partner with MongoDB to help enterprises accurately extract structured insights from visual content such as images, videos and visual documents! ?? We fundamentally believe that VLMs are going to revolutionize the unstructured data and ETL landscape. Our combined solution already enables enterprises to turn their often-untapped unstructured visual content into actionable, queryable business intelligence. #vlm #computervision #etl #mongodb

?? We've been dogfooding VLM Run for our own internal Document AI agents with n8n, and it's been such an eye-opener in how autonomous certain organizations will become in the coming years! ?? Since it's recruiting season for us, we built a fully automated resume parser using our Document -> JSON APIs. It watches our gmail inbox, triggers on PDF resumes being attached, and automatically extracts contact details, GitHub, LinkedIn, and populates a Google Sheet for triaging. Our VLM agents can go even deeper into their Github profiles, Google Scholar etc for deeper analysis of their technical background before we even triage the candidates. ?? I think every enterprise will have their own suite of personalized multi-modal agents running experiments, doing mundane tasks and becoming 10x more productive in their execution. If you're not already automating away the mundane, prepare to be left behind. ?? Parsing Documents Guide: https://lnkd.in/gfMjpxce ?? Document -> JSON API: https://lnkd.in/g9WKVuVa #automation #n8n #hiring #vlm

?? In our latest guide on Image Cataloging, we show how VLMs can simultaneously perform multiple computer-vision tasks such as image captioning, tagging and classification with a custom Pydantic schema! ?? Perfect for e-commerce, visual content management, and more! Check out our image-cataloging example using a sample product dataset from Hugging Face. https://lnkd.in/gPEqMkqS TL;DR ?? Generate captions, tags, and descriptions easily from images ??? Create structured JSON for product catalogs ?? Enable powerful hybrid semantic image search with existing DBs like MongoDB Key features: ? Tailored visual captioning, extraction and classification ? Structured JSON outputs with custom Pydantic schemas ? Scalable to large image catalogs with our fine-tuning API (for enterprises) ?? Interested in trying it out? Sign up for a free trial at VLM Run (https://vlm.run) #ComputerVision #ImageProcessing #AI #ProductCataloging #llm

?? By popular demand, we’ve added OpenAI compatibility to VLM Run! https://lnkd.in/gv6i4mYk ???? This means that with just 2 lines of code-change, you'll be able to access our VLM-1 model directly using the existing OpenAI SDK (JS, Python etc). We think this will be a game-changer for developers who are already using OpenAI and want to quickly take VLM Run for a spin and see the obvious benefits. OpenAI API Compatibility Reference https://lnkd.in/gjWKNdgq API Docs https://docs.vlm.run Let us know what you think! #openai #llm #vlm

?? VLM Run (https://vlm.run) is now live - we’re onboarding a few more customers to our visual AI platform today. {..} If you're looking to extract structured insights (JSON) from rich visual PDFs we’ve added some amazing document AI features into VLM-1 lately that sets us apart from the other vision providers out there. ???Dense table extraction: Accurately extract content from dense tables and distill document summaries into strongly-typed JSON. https://lnkd.in/gt-W3JrK ?? Chart and table grounding: Cite your sources by localizing relevant charts and tables. https://lnkd.in/gihaYnmk ????Fine-tuning: Highly-specialize VLM-1 for accurate document AI in just a few clicks (enterprise-only). https://lnkd.in/gBcyt343 If you’re interested in a free trial, sign-up on the form linked in the comments. #computervision #etl #vlm #llm #foundationmodels

?? We’re excited to announce VLM Run today - platform to help you reliably bring the power of visual intelligence into your applications. ?? We’ve built VLM Run from the ground up, innovating on VLM-1 - our highly-specialized Vision-Language Model (VLM) that excels at accurately extracting JSON from images, videos, documents and even rich presentations. ?????We’ve spoken to developers who’ve tried to incorporate vision into their apps using frontier APIs. Their responses have been eerily similar - these vision models are “just not there yet” for their use-case. ?? We’ve changed that - with VLM Run, our customers have already been able to rapidly operationalize our VLMs into their AI-powered stack. {..} Unlike frontier APIs, our visual models are NOT designed for chat. Instead, we built it with accuracy, latency and the ability to take reliable actions in mind - with the clear vision of enabling reliable visual agentic systems. ???Sign up for VLM Run access today (see comments below)! ?? API docs / Cookbooks also linked (see comments below)! #ai #etl #unstructureddata #documentai #llm #vlm

? Day 1 at #CVPR2024. Need all the caffeine I can get to power through the day. ?? If you’re training VLMs / Visual agents, let’s chat (DM)! Coffee is on me / Autonomi AI!