Pointable

We’re launching 🚀 a new blog series on building real-world LLM agents, starting with HomeworkBot: ⭐ What: An AI to help students track assignments and manage study schedules. ❓ Why: Supports students in organizing and prioritizing amidst online learning challenges. 💡 Insights: We’ll share our development process, including data handling and platform integration. Stay tuned for updates! #llm #ai #edtech #futureoflearning

Agentic AI systems are getting a reality check. It's much harder to deploy reliable, complex agents than people initially thought. There are many reasons why, but keep in mind, any agentic system is comprised of multiple modules, each with many settings / hyper-parameters. At Pointable, we recently dove deep into optimizing a financial Q&A system. Here's what we found: * Embedding choice can make or break performance (we saw a 43% improvement with the right model) * Popular heuristics like "bigger is better" or "always choose the top-ranked model" often fail But here's the kicker: embedding selection is just one of many interacting factors. Especially when building agents to connect to external tools, rely on custom, verified information, and validate user inputs during interactions. The solution? Systematic, end-to-end optimization powered by MLOps principles. We've built Pointable around two key ideas: * Empirical evaluation on real data * Focusing on data quality and comprehensiveness If you're building a RAG system and feeling stuck in the prototype phase, we'd love to help. Our team has guided numerous projects from concept to production-ready applications. Reach out if you're working on this! #MLEngineer #AI #LLMs #EmbeddingModel #RAG #MachineLearning #EnterpriseAI #NLProc

Agentic AI systems are getting a reality check. It's much harder to deploy reliable, complex agents than people initially thought. There are many reasons why, but keep in mind, any agentic system is comprised of multiple modules, each with many settings / hyper-parameters. At Pointable, we recently dove deep into optimizing a financial Q&A system. Here's what we found: * Embedding choice can make or break performance (we saw a 43% improvement with the right model) * Popular heuristics like "bigger is better" or "always choose the top-ranked model" often fail But here's the kicker: embedding selection is just one of many interacting factors. Especially when building agents to connect to external tools, rely on custom, verified information, and validate user inputs during interactions. The solution? Systematic, end-to-end optimization powered by MLOps principles. We've built Pointable around two key ideas: * Empirical evaluation on real data * Focusing on data quality and comprehensiveness If you're building a RAG system and feeling stuck in the prototype phase, we'd love to help. Our team has guided numerous projects from concept to production-ready applications. Reach out if you're working on this! #MLEngineer #AI #LLMs #EmbeddingModel #RAG #MachineLearning #EnterpriseAI #NLProc

🚨 Important read for those building LLM + retrieval systems! 🚨 Our CEO, Andrew Maas, shares some valuable insights on embedding models for similarity in RAG-LLM architectures. Check out our case study in the blog post to learn how to evaluate embedding models properly and optimize end-to-end for better performance. ⚙️💡 👉 Read the blog [https://lnkd.in/eZHepEvy] and dive into the comments to share your thoughts with us! #EnterpriseAI #LLM #Agentic #MLengineering #RAG #NLPRoc

Hot take: LLMs are not going to create artificial general intelligence in the near term. But does that matter? Not if you’re trying to create enterprise value from LLMs over the next few quarters. The existing technology, even if it were frozen today, holds the promise of unlocking hundreds of billions—if not trillions—of dollars in economic value. Large Language Models (LLMs) have reached a pivotal point where they can be leveraged to develop applications and automate various aspects of human work at a relatively low cost. From automating repetitive documentation tasks in enterprise, supporting customers with helpful chatbots, to a new category of consumer-facing AI companions. What’s holding most people back trying to build these systems is a lack of experience and tooling to develop LLM applications like end-to-end complex systems. The evolution is less about creating a super-intelligent entity and more about engineering production-grade solutions through LLMs that are optimized end-to-end for particular tasks. Yes there is engineering and design work required, but you can build applications today we thought were impossible 5 years ago. I recently sat down with Jason Stoughton on the The Pulse of AI podcast’s most recent episode. If you’re building in this space right now, go give it a listen. #AGI #RAG #LLM #enterpriseai #languageai #chatbot #NLProc

Have you ever tried setting up a 💬 chatbot using RAG and found it tricky to get the correct answers? You're not alone! We've been working on 🚀 optimizing RAG systems to make them more accurate and reliable for real-world use. 👀 Check out our blog for more details.

🎙️ Our co-founder & CEO, Andrew Maas, shares insights on AI, entrepreneurship, and Pointable's innovative retrieval systems in the most recent episode of The Pulse of AI 🚀 Tune in for inspiration & knowledge! 🎧

🌟 Our founder, Andrew Maas, recently attended #HWAISummit in Dallas! 🏙️ Andrew was part of an insightful panel discussion on AI use cases in the mortgage industry. 🧠💼 Topics included building your AI team, budgeting for AI, and letting AI alleviate workflow pain points. Check out this sizzle reel where Andrew introduces himself and shares his vision for the future of AI in mortgage! 🎥✨ P.S. Interested in a deeper conversation? Grab some time with Andrew using his office hours link: https://lnkd.in/ezh5yc3M 📅 #AI #MortgageTech #Innovation #TeamBuilding #WorkflowEfficiency #Leadership HousingWire Rick Roque

We are excited to welcome the StartX Summer 2024 cohort! Featuring some amazing Stanford students, professors, and alumni building a better future ☀ Make sure you follow the StartX LinkedIn account as we share some of their stories

It's been a blast 🚀 spending the past few days with the Summer 2024 StartX cohort! A huge thank you to Kyle Wong and the entire StartX team for welcoming our founders, Andrew Maas and Scott Wey. We're excited for what's ahead!