Autoblocks AI

I'm so incredibly excited about the positive feedback we've gotten about Autoblocks AI Launch Week. Our goal is to address the critical yet often overlooked aspects of AI safety and trust in industries like healthcare. Based on the feedback so far, I think we’re on the right track. Here’s a quick recap of features we released: ??? AI Trust Center: A centralized hub for AI performance metrics, security certifications, compliance documentation, and safety controls. ?? Workflows: UI-based AI app/agent builder that empowers non-technical stakeholders to participate in AI development ?? Risk Center: Recurring security and performance testing dashboard that automatically evaluates your applications. Includes built-in "penetration testing" capabilities with a growing library of security tests. ?? Expert Feedback: Clinicians can now provide structured inputs that directly shape model development and improve AI output quality. ?? New Deployment Options: You can now connect your cloud account and deploy Autoblocks to any region in one click, supporting existing or new VPCs. This wasn't just another product update - we introduced substantial features that represent a massive step forward in AI development. And while there's still much work ahead, Launch Week has marked an important milestone in delivering on our vision to building trustworthy AI systems. We're excited to hear your questions and feedback and will continue to ship more features in the coming weeks. Stay tuned! ????

Launch Week Final Day: ?????? ???????????????????? ?????????????? ?? We've completely overhauled our deployment process for Autoblocks, making it easier than ever to set up on your own infrastructure. Our brand new portal allows you to connect your cloud account and deploy Autoblocks with a single click into your cloud provider. Here’s how it works: → Select your cloud account in the portal and paste in your ID to connect your account. → Deploy an Autoblocks instance to any region worldwide, either into an existing VPC or we'll create a new one for you. → Manage your deployment, roll out updates automatically, and streamline the entire process. → Keep your data geographically located (if needed). → For teams with physical isolation requirements, we offer managed Autoblocks in FRA on dedicated compute where we manage it in our cloud infrastructure. This is a game-changer for teams needing more control over their deployments while maintaining ease of use. Check out Adam's demo below and see how these new deployment options can transform your workflow ????

Launch Week Day 4: Expert Feedback ?? Building AI in regulated industries means getting expert input right. But right now, the feedback process is a mess - scattered across tools, hard to track, and difficult to action. That's exactly why we built Expert Feedback with 3 core capabilities: 1?? Centralized feedback collection where experts can comment, score, and override decisions 2?? Organized tracking that links feedback to specific outputs and evaluation categories 3?? Direct pathways to implement improvements through model fine-tuning and workflow optimization With this, clinicians can now provide structured inputs that directly shape model development and improve AI output quality. And this is only just the beginning ???? P.S. If you’re building AI in regulated or high-stakes industries, how are you currently managing expert feedback? Would love to hear your thoughts!

Launch Week Day 3: Autoblocks AI ???????? ???????????? ??? As part of our mission to help enterprises build safer, more reliable AI apps, we're taking application security to the next level (zero setup required). Risk Center continuously monitors your applications and proactively suggests ways to improve security and performance across 4 key areas: 1?? Test case coverage Proactively monitor production activity, identify gaps in testing and get targeted improvement suggestions 2?? PII leak detection Find out where sensitive data might be unintentionally exposed to LLMs or external services—before it becomes a problem. 3?? Prompt injection protection Identify vulnerabilities in your prompts and tackle risks like injection attacks proactively, especially in regulated environments where safety is paramount. 4?? Evaluator coverage It’s not just about what you test but how you measure performance. We ensure comprehensive evaluation using automated tools and human-in-the-loop reviews. With Risk Center, we’re making it easier than ever to manage risks and deploy applications you can trust. Check out Adam's demo below or visit the link in the comments to learn more. Let’s make the future of AI secure ??

Launch Week Day 2: ???????????????????? ??????????????????, a new collaborative way to build safe AI applications. We noticed a growing frustration among teams trying to build AI applications: ? The process was too technical ? Lacks proper safety measures ? Not enough customization options Workflows by Autoblocks is UI-based AI app/agent builder, with a focus on safety & trust. ? Enterprise-level customization that connects with any existing AI system ? Built-in safety testing features that ensure your AI apps work as intended ? A user-friendly interface that??enables non-technical team members to participate in AI development Read more here: https://lnkd.in/eMqf8h9b

To kick off Autoblocks AI Launch Week, we're excited to release ???? ?????????? ????????????: a standardized framework for AI vendors to share AI performance and safety metrics. The AI landscape has a major transparency problem. Vendors are struggling to communicate their performance metrics in a way that makes sense to prospects. And customers are drowning in whitepapers, unable to properly evaluate different AI solutions against each other. That's why we built AI Trust Center. It plugs directly into any AI system (whether built in Autoblocks or not) to help companies standardize how they present their AI capabilities. Here's what you get: ? Comprehensive test results and performance metrics ? Real-time updates on evaluation metrics ? Relevant industry certifications ? Detailed FAQs about methodology We’re making this completely free because we believe transparency shouldn't come with a price tag. Join our beta waitlist (link in comments) or stay tuned for general availability by end of Q1.

?? Big things are coming… Autoblock’s first Launch Week kicks off Jan 13! 5 days. 5 launches. Join us as we transform how you: ? Identify AI application vulnerabilities ? Detect emerging edge cases ? Integrate safety and trustworthiness Stay tuned. It’s going to be a week to remember.

Discover how to enhance the accuracy and dependability of LLM judges. Haroon Choudery delves into strategies for harnessing human expertise to refine LLM evaluations. Learn all about fine-tuning evaluators, utilizing structured feedback loops, and aligning LLM judgments with human preferences ??

Fine-tuning is one of the most effective ways to enhance an LLM's performance for specific tasks. In this webinar, we'll demonstrate how to fine-tune a base model using the Together Python Library and compare its performance vs. a base model using Autoblocks AI's testing and evaluation solution.

Join us for an insightful webinar featuring Together AI and Autoblocks AI this Wednesday! ?? We'll be comparing fine-tuned vs. base AI models with experts Hassan El Mghari and Adam Nolte. ?? Register now: https://lu.ma/x6v3wpv6