<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Rick Bhardwaj on Medium]]></title>
        <description><![CDATA[Stories by Rick Bhardwaj on Medium]]></description>
        <link>https://medium.com/@rick_27144?source=rss-a7cef8cf3d0f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*00T_2aEfISN7yyR6</url>
            <title>Stories by Rick Bhardwaj on Medium</title>
            <link>https://medium.com/@rick_27144?source=rss-a7cef8cf3d0f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 12:59:43 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@rick_27144/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[We Free Professionals from Bad Software]]></title>
            <link>https://blog.strangelooplabs.ai/we-destroy-software-monopolies-90fde9b8532c?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/90fde9b8532c</guid>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Mon, 12 May 2025 15:35:16 GMT</pubDate>
            <atom:updated>2025-05-15T17:48:16.983Z</atom:updated>
            <content:encoded><![CDATA[<p>The systems that govern the world — healthcare, payroll, insurance, taxes, commerce, finance — are all powered by legacy software built in the COBOL era. I call these software companies the<strong> Legacy Software Cartel:</strong></p><ul><li>100% of F100 companies use Reuters for corp tax</li><li>94% of hospitals use the same two EHR vendors.</li><li>99% of loans are served by CoreLogic.</li><li>87% of global commerce is run by SAP.</li><li>1 in 6 Americans paychecks is processed by ADP.</li><li>95% of retail deposits sit on FIS and Fiserve mainframes.</li></ul><p>These systems of records aren’t going anywhere, but trying to customize these behemoths to your organization is often <strong>the most expensive IT projects in history</strong>, bogging down customers in year-long implementation projects and endless “it’s on the roadmap” promises.</p><p>So the world’s most important organizations have to settle for throwing professionals at one-size-fits-all software products.</p><p>Our smartest professionals are forced to become the human glue between systems and clients, chained to their mouse and keyboard, clicking through MS-DOS-style menus and manually entering data, occasionally glancing at their degree on the wall and wondering where it all went so wrong.</p><p>These “systems of records” aren’t going anywhere.</p><p>This insane arrangement exists because A. These “systems of records” aren’t going anywhere and B. every organization’s workflow on top of these systems are unique, and creating custom software to automate them is <strong>expensive and slow.</strong> Customization projects are years-long implementation projects — or “it’s on the roadmap” promises pay off.</p><p>So the world’s most important organizations have to settle for throwing professionals at one-size-fits-all software products. Until now.</p><h3>The Era of Truly Personalized Computing</h3><p>Here’s how Strange Loop frees professionals from the Cartel</p><ol><li>We go into complex, regulated industries, sit with their professionals to understand their processes and deploy systems perfectly tailored to their workflows.</li><li>We’re able to deploy workflows in <strong>weeks </strong>and integrate into any system you already use. No rip-and-replace required, no new tools to learn. <strong>Just faster, better outcomes without all the clicking.</strong></li></ol><p><strong>Wait, how does this scale?</strong></p><p>This is now possible because AI has fundamentally transformed software creation:</p><p><strong>Strange Loop </strong>delivers perfectly tailored software workflows in weeks, without forcing you to rip-and-replace legacy systems or suffer through another one-size-fits-all tool.</p><p>AI has fundamentally changed how software is built and distributed.</p><p>What makes this possible is two products at Here at Strange Loop, we’ve built:</p><ul><li><strong>A core AI platform</strong> that process unstructured data, navigate complex regulations, all hosted on secure, scalable infrastructure that can integrate into any system.</li><li><strong>Proprietary AI coding tools </strong>that enable our engineers to configure our platform at unprecedented speed and costs.</li></ul><p>We go into complex, regulated industries, sit with their professionals to understand their processes and deploy systems perfectly tailored to their workflows.</p><p>We’re able to deploy workflows in <strong>weeks </strong>and integrate into any system you already use. No rip-and-replace required, no new tools to learn. <strong>Just faster, better outcomes without all the clicking.</strong></p><p>This approach at the price and timeline customers want is only recently possible.</p><p>Customers love this approach:</p><blockquote><em>“The people who are working with me are literally jumping for joy. This is the cleanest data they’ve ever seen.”</em></blockquote><blockquote><em>“This feature made my month. I should have watched this video in the throes of busy season to know help was on the way!”</em></blockquote><p>I know — this sounds fake. But you have to realize that our customers have never experienced purpose-built software tools before — only rigid tools built for someone else’s process.Every mass market software product is, by definition, a compromise.</p><p>Software was supposed to be the greatest productivity unlock in human history. We still believe it can be. <strong>Bad software delende est.</strong></p><h3>Who We Want to Work With</h3><p>In a world where software gets cheaper to develop, the real bottleneck is smart systems thinkers who can parachute into complex organizations, understand their problems, deploy solutions, and contribute learnings back to our platform. It’s closer to a startup founder than a traditional engineering role — you’ll wear engineering, PM, and sales hats. If that sounds interesting to you, get in touch with us <a href="https://jobs.ashbyhq.com/strange-loop-labs">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=90fde9b8532c" width="1" height="1" alt=""><hr><p><a href="https://blog.strangelooplabs.ai/we-destroy-software-monopolies-90fde9b8532c">We Free Professionals from Bad Software</a> was originally published in <a href="https://blog.strangelooplabs.ai/">Strange Loop Labs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[So why did we pick Strange Loop Labs as our new name?]]></title>
            <link>https://medium.com/@rick_27144/so-why-did-we-pick-strange-loop-labs-as-our-new-name-346269cdb86c?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/346269cdb86c</guid>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Tue, 04 Feb 2025 16:04:23 GMT</pubDate>
            <atom:updated>2025-02-04T16:04:53.677Z</atom:updated>
            <content:encoded><![CDATA[<h3>Why are we named <a href="https://www.linkedin.com/company/strange-loop-labs/">Strange Loop Labs</a>?</h3><p>A strange loop is a system that paradoxically ends up where it started. Think of M.C. Escher’s “Drawing Hands” — each hand drawing the other into existence.</p><p>It seems like we might be watching a strange loop emerge in real-time. Software is (almost) able to write itself. This is very weird. Every assumption about how software is built, sold, and marketed might be outdated. SaaS made sense…in an era where software was expensive to create.</p><p>We want to build one of the first companies designed around autonomous software agents from the start.</p><p>So we put a reminder into <a href="https://www.strangelooplabs.ai/">our name</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*4zj6tqYu9BSbWq13kwlVcw.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=346269cdb86c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Announcing Strange Loop Labs]]></title>
            <link>https://blog.strangelooplabs.ai/announcing-strange-loop-labs-26d73117973c?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/26d73117973c</guid>
            <category><![CDATA[fundraising]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Thu, 23 Jan 2025 16:01:12 GMT</pubDate>
            <atom:updated>2025-05-09T01:12:00.308Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*t0AL7Uvjh-rxt5bihaCybA.png" /></figure><h3><strong>AI should provide outcomes, not tools.</strong></h3><p>As much as we love our chatbots, it is ultimately just another tool for professionals to wrangle on top of their 20+ other SaaS tools.</p><p>We believe the real AI opportunity in <strong>automating</strong> <strong>processes, end to end, at the world’s most complex regulated industries: </strong>tax, finance, law, healthcare, insurance, government, and more.</p><p>We work alongside professionals and use our platform to process their messiest data and reasons how to solve their hardest problems. Customers don’t learn new tools or attend onboardings — they just see <strong>faster, better outcomes inside the tools they already use.</strong></p><p>Over the past year, we have reached over 7 figures of ARR, <a href="https://www.linkedin.com/feed/update/urn:li:activity:7288253493511954433/">raised a seed round</a> from incredible AI investors, and we’re <a href="https://jobs.ashbyhq.com/strange-loop-labs">building a team</a> where <strong>each engineer creates $100M of value</strong> for our customers.</p><p>We believe that will be achievable with the right tooling and team — and we believe Strange Loop will become the most valuable software company in the world as a result.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=26d73117973c" width="1" height="1" alt=""><hr><p><a href="https://blog.strangelooplabs.ai/announcing-strange-loop-labs-26d73117973c">Announcing Strange Loop Labs</a> was originally published in <a href="https://blog.strangelooplabs.ai/">Strange Loop Labs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My advice to enterprises trying wrap their heads around their AI strategy]]></title>
            <link>https://blog.strangelooplabs.ai/my-advice-to-enterprises-regarding-ai-strategy-1a08f7145ab4?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/1a08f7145ab4</guid>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Sun, 11 Aug 2024 19:44:48 GMT</pubDate>
            <atom:updated>2024-08-11T23:09:29.996Z</atom:updated>
            <content:encoded><![CDATA[<ul><li>Don’t get too hung up on chatbots. The easy wins are turning unstructured data into structured data.</li><li>Find the messiest data problem at your company. That’s your gold mine. (Hint: it’s probably a pile of documents no one wants to deal with.)</li><li>Build an AI pipeline around it. Make it spit out clean data into your existing tools: Excel, Salesforce, etc.</li><li>Please, please, please build tools to audit the outputs.</li><li>Your ops team will thank you. You’ll save your first $1M before you know it.</li><li>Chatbot use cases will naturally arise once you do this.</li></ul><p>Don’t overthink it. All enterprise software value has come from automating business processes by reading, transforming, and storing data. 80% of data was too unstructured for software to handle. Now, with LLMs, it isn’t.</p><p>Imagine you’re a NYC real estate developer and woke up to find out Manhattan is 10x bigger than it was yesterday. That’s what’s happening to enterprise software right now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yKbFHLMfZRFS5IJoZwl1lA.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1a08f7145ab4" width="1" height="1" alt=""><hr><p><a href="https://blog.strangelooplabs.ai/my-advice-to-enterprises-regarding-ai-strategy-1a08f7145ab4">My advice to enterprises trying wrap their heads around their AI strategy</a> was originally published in <a href="https://blog.strangelooplabs.ai/">Strange Loop Labs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Accountants use Guru AI Workflows]]></title>
            <link>https://blog.strangelooplabs.ai/how-accountants-use-guru-ai-workflows-0abd4701b9b5?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/0abd4701b9b5</guid>
            <category><![CDATA[agents]]></category>
            <category><![CDATA[ai-accounting]]></category>
            <category><![CDATA[ai-workflow]]></category>
            <category><![CDATA[accounting]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Wed, 10 Jul 2024 14:48:23 GMT</pubDate>
            <atom:updated>2024-07-24T18:36:03.526Z</atom:updated>
            <content:encoded><![CDATA[<p>Guru is a cloud AI platform for automating complex document workflows at massive scale. <strong>With Guru AI Workflows, tax returns that took 2+ hours are now completable in 20 minutes. </strong>By the end of the 2024 tax season, Guru will have processed millions of tax returns, saving CPAs thousands of hours. Let’s dive deeper.</p><h4>Death, Taxes, and AI</h4><p>𝐊-𝟏𝐬 are the forms used to file state income taxes. If you’re not an accountant, you probably think they’re some obscure form you hope you filed last year. If you are an accountant, you just felt your blood pressure spike.</p><p>In the world of accounting, few tasks are as daunting as processing state K-1s. Traditional automation tools like OCR/RPA have failed to crack this process because K-1s come in an endless assortment of complex formats.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8STnSe1DZPl0SnVYmf761Q.png" /></figure><p>However, we’ve found that new generative AI models, combined with AI-native review tools, can finally generalize to these documents and speed up accountant workflows.</p><p>Guru AI Workflows involve three components:</p><h4><strong>1. Document Engine</strong></h4><p>Guru’s proprietary visual document models, trained on vast datasets of K-1 forms, can “see” tax documents as an accountant does, understanding complex layouts and summarize information leveraging the AI-powered tax analysis.</p><h4><strong>2. Human Review</strong></h4><p>Critical business workflows can’t be fully handed off to AI (yet.) We’ve built visual tools that let humans quickly audit AI decisions and earn trust.</p><h4><strong>3. Universal Integration</strong></h4><p>Guru seamlessly connects with any tax software like CCH, Reuters, and even Excel, and also exposes APIs your developers can integrate with.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OJkFGyaoNOwj_hCP7GcxQg.png" /></figure><p>While chatbots dominate headlines, it’s these behind-the-scenes AI workflows that are delivering tangible ROI to accounting firms today. The ROI isn’t just in dollars (though there are plenty of those), but in freeing CPAs from mind-numbing data entry to focus on more creative, strategic, fulfulling tasks that attracted them to accounting in the first place.</p><h4>The robots are here, and they’re ready to do your taxes.</h4><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0abd4701b9b5" width="1" height="1" alt=""><hr><p><a href="https://blog.strangelooplabs.ai/how-accountants-use-guru-ai-workflows-0abd4701b9b5">How Accountants use Guru AI Workflows</a> was originally published in <a href="https://blog.strangelooplabs.ai/">Strange Loop Labs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing Guru AI Workflows]]></title>
            <link>https://blog.strangelooplabs.ai/introducing-guru-ai-workflows-220f9d27a29b?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/220f9d27a29b</guid>
            <category><![CDATA[agents]]></category>
            <category><![CDATA[serverless]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-workflow-automation]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Tue, 25 Jun 2024 17:16:11 GMT</pubDate>
            <atom:updated>2024-08-02T18:08:37.183Z</atom:updated>
            <content:encoded><![CDATA[<h4>The Next Phase of Enterprise AI</h4><p>Today we’re announcing <strong>Guru AI Workflows</strong>, a cloud AI platform for automating business processes with complex documents and data.</p><p>Chatbots are just the beginning. Soon, every enterprise will run on superintelligent systems working behind-the-scenes to analyze data and make decisions millions of times a year. Customers won’t see it, but they’ll feel it as business processes become smarter and smoother everywhere. We call these systems <strong>AI Workflows.</strong></p><p>Our team previously built AI platforms for Amazon Alexa and AWS AI, helping write the playbook on serving and scaling AI workflows. Our goal with <strong>Guru AI Workflows </strong>is to help enterprises focus on transforming business operations, not AI infrastructure.</p><h3>Here’s how customers are using <strong>AI Workflows</strong> today</h3><h4>AI Tax Preparers</h4><p><strong>F100 accounting firms</strong> are applying <strong>AI Workflows</strong> to tax prep. Accountants manually review millions of documents, leading to a staggering <a href="https://vintti.com/blog/according-to-a-study-99-of-accountants-suffer-from-burnout/"><em>99%</em> <em>burnout rate</em>.</a> <strong>AI Workflows</strong> are saving accountants are saving <strong>hours per tax return</strong> by parsing key details from unstructured tax forms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*xmHXVLBG8fJqprivePYSEA.gif" /></figure><h4>AI Insurance Underwriters</h4><p><strong>Modern insurance firms like </strong><a href="https://www.decent.com/"><strong>Decent</strong></a> are applying <strong>AI Workflows</strong> to policy comparison. Insurance brokers manually review dozens of insurance quotes for every customer every year. With <strong>AI Workflows</strong>, brokers are saving hours per quote and highlighting policy discrepancies humans often miss.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nF-ceId6soLCWL48QA-6Gg.gif" /></figure><h3>Capabilities</h3><h4>Secure, Scalable AI Infrastructure</h4><p>Guru AI Workflows orchestrate <strong>proprietary task-specific models</strong> and <strong>foundation models</strong> into an execution graph. Our task-specific models (i.e., New York state tax analysis) are optimized to run on our serverless, stateless cloud infrastructure, scaling to millions of documents without storing any data and costing pennies per page. For more general tasks (i.e., summarization), we benchmark and dynamically select the best foundation model API for the task.</p><h4>Analyze Any Document from Any System</h4><p>Guru can understand and process your most complex documents <strong>in</strong> <strong>any format</strong> and <strong>from any system,</strong> including emails from that ancient PDF scanner in your office.</p><h4>Go Beyond the Chatbox</h4><p>Critical business workflows can’t be fully handed off to AI (yet.) We’ve built visual tools that let humans quickly audit AI decisions and earn trust.</p><h3>Early Access</h3><p>This has been by far our highest volume product so we’re onboarding customers in batches. If your organization handles thousands (or millions) of documents annually and is interested in applying AI to streamline workflows, get in touch with us <a href="https://tally.so/r/mDd5Kp">here</a>.</p><h3>About Guru</h3><p>We’re an applied multimodal AI team building AI tools for enterprises. As we gear up for our public launch, we’re selectively adding to our founding team. If you’re passionate about multimodal AI and obsessed with solving complex, real-world business challenges, get in touch with us <a href="https://www.notion.so/Founding-AI-Engineer-SWE-8e0b406dc8cc488e8ce658f4158f634b?pvs=21">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=220f9d27a29b" width="1" height="1" alt=""><hr><p><a href="https://blog.strangelooplabs.ai/introducing-guru-ai-workflows-220f9d27a29b">Introducing Guru AI Workflows</a> was originally published in <a href="https://blog.strangelooplabs.ai/">Strange Loop Labs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Guru Completes AWS Foundational Technical Review for the Guru AI Video Analysis Platform]]></title>
            <link>https://blog.strangelooplabs.ai/guru-completes-aws-foundational-technical-review-for-the-guru-ai-video-analysis-platform-e8d585d4e2bd?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/e8d585d4e2bd</guid>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Wed, 29 Nov 2023 17:55:11 GMT</pubDate>
            <atom:updated>2023-12-06T19:17:25.297Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4JOHIb-gDOjg9sZrpKyDug.jpeg" /></figure><p><strong>New York , New York (November 2023) —</strong> Guru, an AI video analysis developer platform and an AWS Partner Network (APN) member, has successfully completed the <a href="https://aws.amazon.com/partners/foundational-technical-review/">AWS Foundational Technical Review</a> (FTR) for its AI video analysis platform. Guru makes it simple to build complex AI video analysis applications such as analyzing tennis serve form, identifying defects on a conveyor belt, or tracking the average wait time in a checkout line. With Guru, users can create applications that leverage complex vision models in a few lines of code and deploy to any runtime with one-click deployment.</p><p>The AWS FTR allows AWS Partners like Guru to identify and remediate risks in their software or solutions. Whether your software is delivered as a SaaS solution or deployed by customers, the FTR helps AWS Partners identify AWS Well-Architected best practices specific to their software. Service offering FTRs assess an AWS Partner’s Service Offering against a specific set of best practices most critical for customer success.</p><p>Guru’s collaboration with AWS enables developers to leverage the power and flexibility of the cloud to train, develop, and deploy AI video models and scale their AI video analysis applications to handle millions of hours of video. This streamlines the process of developing and deploying AI video analysis solutions, making it more accessible to a broader range of developers and applications.</p><p>Rick Bhardwaj, Chief Executive Officer for Guru, commented “We are thrilled to partner with AWS in enhancing our AI video analysis solutions. At Guru, we are dedicated to providing our customers with cutting-edge solutions that harness the power of AI and cloud computing, enabling them to analyze video data efficiently and accurately. Our collaboration with AWS will help us deliver exceptional solutions to our customers.”</p><p><strong>Press Contact:</strong></p><p>Rick Bhardwaj</p><p>CEO, Guru</p><p>(917) 391–0754</p><p><a href="/service/mailto:rick@getguru.ai">rick@getguru.ai</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e8d585d4e2bd" width="1" height="1" alt=""><hr><p><a href="https://blog.strangelooplabs.ai/guru-completes-aws-foundational-technical-review-for-the-guru-ai-video-analysis-platform-e8d585d4e2bd">Guru Completes AWS Foundational Technical Review for the Guru AI Video Analysis Platform</a> was originally published in <a href="https://blog.strangelooplabs.ai/">Strange Loop Labs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Guru Console: An IDE for video AI]]></title>
            <link>https://medium.com/@rick_27144/guru-console-an-ide-for-video-ai-5e5d6dbb91a8?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e5d6dbb91a8</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[development]]></category>
            <category><![CDATA[computer-vision]]></category>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Thu, 11 May 2023 16:40:38 GMT</pubDate>
            <atom:updated>2023-05-11T16:45:10.886Z</atom:updated>
            <content:encoded><![CDATA[<h4>Today, we’re opening up access to the Guru Console, a web-based tool for creating AI models that understand complex video concepts.</h4><p>Powerful vision models are enabling a new generation of apps in healthcare, sports, content search, security, and more. But we’ve found that domain experts have difficulty applying AI-powered analysis to video content:</p><ul><li>High-level video concepts like “who ran the fastest in this video?” require chaining together multiple complex models.</li><li>Pretrained models don’t generalize well to real user videos, but developers struggle to finetune models without tooling.</li></ul><p>We’ve been working closely with our early developers to build intuitive visual tools to overcome these bottlenecks. We’ve packaged these tools into the <strong>Guru Console</strong>, a web-based tool for creating AI models that understand complex video concepts. Guru Console users have built models that:</p><ul><li>Perform custom physical therapy assessments</li><li>Analyze the running form of the top NFL draft prospects.</li><li>Execute semantic search on video streams</li></ul><p>Let’s walk through how to use the Guru Console to visually debug AI models, teach Guru new concepts, and get onboarded onto the console.</p><h3><strong>See What Guru Sees</strong></h3><p>When you upload a video to Guru, you will see a visual representation of Guru’s AI-powered understanding of your video.</p><p>We’ve broken these down into three major “video concepts”:</p><ul><li><strong>Visual Recognition:</strong> Recognizing and tracking people, objects, and environments in the video</li><li><strong>Video Understanding:</strong> Understanding actions and scenes Guru, like a baseball swing or picking up a box</li><li><strong>Analysis:</strong> Timeseries analysis on what Guru recognizes — i.e., rep counting, form scoring, video trimming.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tVuSPYmrrzDG82NMG-i-ug.png" /></figure><h3><strong>Teach Guru New Video Concepts</strong></h3><p>From this same screen, you can seamlessly <strong><em>teach</em></strong> Guru new, specialized video concepts or improve Guru’s accuracy on new video datasets. The Console will guide you through both. Just click edit next to whichever concept you want to create or edit.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VaJiMezWXW8K3l_odmQNYw.png" /></figure><p>Once you’re done and save, Guru will use your annotations to improve the models using an ML process called <strong>finetuning</strong>. This helps with common vision model failures like occluded body parts, limbs out of frame, or new camera angles.</p><p>Typically we see major accuracy gains after a 10–25 videos (and just 1–3 frames in each video.) Here’s an example of the accuracy boost after ~12 videos:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.loom.com%2Fembed%2Fe478305c89cc4dbeab8c45d029f1a85b&amp;display_name=Loom&amp;url=https%3A%2F%2Fwww.loom.com%2Fshare%2Fe478305c89cc4dbeab8c45d029f1a85b&amp;image=https%3A%2F%2Fcdn.loom.com%2Fsessions%2Fthumbnails%2Fe478305c89cc4dbeab8c45d029f1a85b-00001.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=loom" width="1440" height="1080" frameborder="0" scrolling="no"><a href="https://medium.com/media/00c4bebff00727c486bd87853d20097d/href">https://medium.com/media/00c4bebff00727c486bd87853d20097d/href</a></iframe><p>When your model is ready, head over to API Deployment tab to deploy to a hosted API or to an <a href="https://blog.getguru.ai/building-the-state-of-the-art-video-ai-library-for-android-71d2ff6a182f">on-device SDK</a>. Over time, Guru will automatically populate the console with any videos uploaded to the Guru API with your API key. You can finetune your model with any of these videos.</p><h3>Get Started</h3><p>We’re letting in developers after a quick intro call. Get started by emailing support@getguru.ai or signing up <a href="https://2o7cxaf2plv.typeform.com/to/YpRQhslY">here</a>.</p><p>Using the Guru Console requires a Guru subscription, but we do have a developer tier with a free trial.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5e5d6dbb91a8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Guru Live: Building the state-of-the-art video AI library for Android]]></title>
            <link>https://medium.com/@rick_27144/building-the-state-of-the-art-video-ai-library-for-android-71d2ff6a182f?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/71d2ff6a182f</guid>
            <category><![CDATA[android]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Thu, 27 Apr 2023 18:43:09 GMT</pubDate>
            <atom:updated>2023-04-27T18:50:10.290Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/443/1*mawAMbsipgjNNPque-mkbw.gif" /></figure><h4>Today, we’re launching Guru Live for Android. Guru developers can now export their custom Guru models to any Android device in one-click.</h4><p>Based on our benchmarks, Guru Live offers <strong>state-of-the-art accuracy and speed for Android on-device inference, </strong>a notoriously difficult platform for custom AI deployments.</p><p>With this release and our pre-existing iOS support, Guru developers can now create customized, real-time video AI apps that run on virtually any mobile device.</p><h4>Why did we build this?</h4><p>Guru’s hosted <a href="https://docs.getguru.fitness/#introduction">inference APIs</a> is suitable for many video AI use cases, but due to hard constraints around streaming latency and privacy, most real-time video AI use cases will have to run on edge devices like phones.</p><p>However, running video AI on mobile devices has been challenging due to limited processing power. This is especially true for Android devices with varying hardware configurations and limited support for hardware acceleration.</p><p>As a result, most Android developers today settle for pre-trained, optimized vision libraries like <a href="https://www.tensorflow.org/hub/tutorials/movenet">MoveNet</a>, which only supports a narrow set of use cases.</p><p>Guru developers wanted to achieve blazing-fast inference speed on Android devices while retaining the custom vision models they built with Guru.</p><h4>Our Solution + Benchmarking</h4><p>To support this, we built a pipeline that takes a Guru model and optimizes it for memory-constrained ARM devices. The key metrics we focused on are:</p><ul><li><strong>Accuracy,</strong> measured by Average Precision (AP) — the accuracy of inferred keypoints vs. ground truth)</li><li><strong>Speed,</strong> measured by the number of frames per second (FPS) that inferences runs.</li></ul><p>Our customers have told us that <strong>10+ FPS</strong> is required for realtime responsiveness and <strong>65%+ AP</strong> is required for quality motion analysis.</p><p>For benchmarking, we used a low-powered, ~$100 Android device and compared our optimized model against the most popular Android video models: MoveNet-Thunder (accuracy-optimized) and MoveNet-Lightening (latency-optimized.) We compared it to a test-split for a physical therapy movement our customers finetuned their Guru models for.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oKdrbUMM_IRcxRmqBdxwOA.png" /></figure><p>Takeaways:</p><ul><li><strong>We’re significantly faster and more accurate than MoveNet-Thunder.</strong></li><li><strong>We’re roughly as fast as MoveNet-Lightning on CPU, but significantly more accurate.</strong></li></ul><h4>Try It Out</h4><p>To try out Guru live, email me at <a href="/service/mailto:support@getguru.ai">rick@getguru.ai</a> or join the <a href="https://join.slack.com/share/enQtNTEzMDQ5ODUyMzg5MS1iYWNhMmMxYWU3MmRlNmVhOTMwMTlkODQxMGY3ZDhmZDA0YmVkZjdlZjFjMTQ5YTA3MTM2MWNlM2Q4OTM5ZDJk">Guru Developer Slack</a> and DM us. Here you can use Guru’s pre-trained models or create a new vision model in the console and export an Android library for your model.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=71d2ff6a182f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Guru Sports: Powering the Top Prospects at the 2023 NFL Draft]]></title>
            <link>https://medium.com/@rick_27144/guru-sports-powering-the-top-prospects-at-the-2023-nfl-draft-e8aa8c5f486c?source=rss-a7cef8cf3d0f------2</link>
            <guid isPermaLink="false">https://medium.com/p/e8aa8c5f486c</guid>
            <category><![CDATA[nfl-combine]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[nfl-draft]]></category>
            <dc:creator><![CDATA[Rick Bhardwaj]]></dc:creator>
            <pubDate>Thu, 23 Feb 2023 19:50:35 GMT</pubDate>
            <atom:updated>2023-03-06T03:28:33.450Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/932/1*clIz-o1BY2Nqr5eva7qo9Q.png" /></figure><h4>We’re launching Guru Sports, an AI-enabled toolkit to provide elite coaching feedback on mobile video.</h4><p>Check out who’s using it and how it works below, and reach out to support@getguru.fitness to get onboarded!</p><h3>Who’s Using It</h3><p><a href="https://twitter.com/les7spellman/status/1569802871885156353">Spellman Performance</a> is using the Guru Sports for<strong> </strong><a href="https://www.youtube.com/watch?v=L7jhgPrimfM"><strong>speed training</strong></a><strong> half of the projected first round of the 2023 NFL Draft</strong>. Next week, these athletes will put their speed training to work at the NFL Combine 40 yard dash.</p><h3>Les Spellman on Twitter: &quot;Been quietly working on a project with my guys @getgurufitness creating sprint profiles from iphone footage. It&#39;s a game changer. This particular analysis is looking at total thigh range (with flexion/extension angles) &amp; GT/ATresearch cred: @KenClarkSpeed pic.twitter.com/DoxG59Mig1 / Twitter&quot;</h3><p>Been quietly working on a project with my guys @getgurufitness creating sprint profiles from iphone footage. It&#39;s a game changer. This particular analysis is looking at total thigh range (with flexion/extension angles) &amp; GT/ATresearch cred: @KenClarkSpeed pic.twitter.com/DoxG59Mig1</p><p><a href="https://www.breakawaydata.com/">Breakaway Data</a> is using Guru Sports to bring speed training to youth athletes. Over the next few weeks, Breakaway aims to roll this out <strong>to 10,000+ youth athletes.</strong></p><h3>BreakAway Data on Twitter: &quot;And we LOVE it. Use it to measure your sprinting speed in MPH.Simple as filming your run and uploading it in the BreakAway app. We hit you back with verified results in just minutesBreakAway app: https://t.co/dmecenzhs5#athletes #speed #athletetraining #gofast #sports https://t.co/PBXoW4DnAI pic.twitter.com/DaO9cvLIWn / Twitter&quot;</h3><p>And we LOVE it. Use it to measure your sprinting speed in MPH.Simple as filming your run and uploading it in the BreakAway app. We hit you back with verified results in just minutesBreakAway app: https://t.co/dmecenzhs5#athletes #speed #athletetraining #gofast #sports https://t.co/PBXoW4DnAI pic.twitter.com/DaO9cvLIWn</p><p>Reach out to support@getguru.fitness if you want to be next!</p><h3>How It Works</h3><p>Elite sports coaching today relies extensively on film review. This traditionally requires hours of manual coach review and expensive equipment. We’ve talked to many developers who want to automate this process with AI, but this is one of the trickiest usecases in computer vision.</p><p>For example, here’s a typical sprint video (featuring Kentucky QB Will Levis.)</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.loom.com%2Fembed%2F896abd19c5aa4d17afac0645a38a9a7d&amp;display_name=Loom&amp;url=https%3A%2F%2Fwww.loom.com%2Fshare%2F896abd19c5aa4d17afac0645a38a9a7d&amp;image=https%3A%2F%2Fcdn.loom.com%2Fsessions%2Fthumbnails%2F896abd19c5aa4d17afac0645a38a9a7d-00001.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=loom" width="3840" height="2880" frameborder="0" scrolling="no"><a href="https://medium.com/media/629e957865158c6acfaaf283e4cf182d/href">https://medium.com/media/629e957865158c6acfaaf283e4cf182d/href</a></iframe><p>Extracting AI insights on this kind of video requires:</p><ul><li><strong>Accurate, high frame rate motion capture on mobile-quality video</strong></li><li><strong>Automatically recognizing custom sports movements</strong></li><li><strong>Understanding objects and distances on the field</strong></li></ul><p>With the <strong>Guru Sports,</strong> developer can build AI-powered sports analysis models that work on mobile video. For example, we used Guru Sports to add the <strong>Sprints API, which </strong>allows developers to:</p><ul><li>Detect every stride and do precise biomechanical analysis:</li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.loom.com%2Fembed%2Ff010955483294298a26507dd0b0adb52&amp;display_name=Loom&amp;url=https%3A%2F%2Fwww.loom.com%2Fshare%2Ff010955483294298a26507dd0b0adb52&amp;image=https%3A%2F%2Fcdn.loom.com%2Fsessions%2Fthumbnails%2Ff010955483294298a26507dd0b0adb52-1663098665253.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=loom" width="1920" height="1440" frameborder="0" scrolling="no"><a href="https://medium.com/media/4a3811f088dca9312e59adcb214d9276/href">https://medium.com/media/4a3811f088dca9312e59adcb214d9276/href</a></iframe><ul><li>Calculate speed and distance without requiring any equipment — Guru just uses yard-markers and cones on the field.</li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.loom.com%2Fembed%2F44b0d20818ff4ab98f6778d1db4e1ce8&amp;display_name=Loom&amp;url=https%3A%2F%2Fwww.loom.com%2Fshare%2F44b0d20818ff4ab98f6778d1db4e1ce8&amp;image=https%3A%2F%2Fcdn.loom.com%2Fsessions%2Fthumbnails%2F44b0d20818ff4ab98f6778d1db4e1ce8-00001.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=loom" width="1920" height="1440" frameborder="0" scrolling="no"><a href="https://medium.com/media/f98a4042f304ef35760de56d519552b3/href">https://medium.com/media/f98a4042f304ef35760de56d519552b3/href</a></iframe><p>With these platform improvements, developers can quickly add many more sports movements to Guru. If you have a sports movement you’d like to build with Guru Sports, we’d love to help — reach out to support@getguru.fitness.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e8aa8c5f486c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>