Humanloop

If you're leading AI teams, this episode is a must-listen. Peter Gostev shares his experience developing AI strategy previously at NatWest and now Moonpig. His perspective on both big enterprise and startup environments is gold. Learn: 1. How to balance quick wins and experimental projects in your AI strategy 2. Practical ways AI is being used at Moonpig 3. The challenges of deploying AI in a regulated environment 4. Why automating call centres might still be too expensive 🎙️ Tune in: → Youtube https://lnkd.in/eim7uTJE   → Spotify https://lnkd.in/euSYwKBV → Apple https://lnkd.in/e2wJ8FER Some key takeaways: 1. Have a two-pronged approach to AI strategy. First, spend time experimenting with AI models to develop intuition about their capabilities. Second, maintain a balanced portfolio of AI projects, including both quick wins to demonstrate immediate value and larger, more experimental projects for long-term potential. 2. AI doesn't just boost productivity - it enables entirely new types of work. Moonpig improved customer service agent efficiency 15-20% by using AI. They’re also using visual models to tag greeting cards for improved search and widely adopting ChatGPT across the company. Don’t forget to check out our past episodes humanloop.com/podcast if you’re looking to get even more insights into building with AI.

Humanloop NYC team offisite! Working at Humanloop offers tons of fantastic benefits(working on cutting edge AI product is one of them). However, one benefit I'm truly grateful for - team retreats 🇺🇸! We are a distributed team, and nothing builds up trust like train hackathons, cooking pickle pasta sauce and campfire chats 😌 We absolutely not a cul.. Loop Loop Loop Loop!

The Ex-Coinbase CPO wants to build fully AI Employees. I was initially sceptical but by the end of this week's episode of High Agency, I was convinced! Surojit Chatterjee has an amazing background. He launched Google’s mobile search ads and grew it to $50B+ in revenue, was the CPO at Coinbase and now he’s creating universal AI employees. Learn: 1. Why most enterprise AI projects fail (and how to succeed) 2. AI employee use cases in enterprises 3. Challenges with building AI agents 4. Strategies for eliminating AI hallucinations at scale 5. How the business model of SaaS will change 🎙️Tune in: → Youtube https://lnkd.in/dZJ-k8SU → Spotify https://lnkd.in/dbg6Uuni → Apple https://lnkd.in/dVwsB-4b Some takeaways: • Building AI agents presents unique challenges because their performance can fluctuate over time. Unlike traditional software, AI models require continuous monitoring, feedback, and updates to maintain and improve performance. “If you buy any regular SaaS app, it just works the same way unless somebody pushes new code on the back end. But that’s not the case here,” says Surojit. • To reduce hallucinations at scale, Emma uses a “mixture of experts” approach, combining multiple foundation models (both open-source and proprietary) to cross-reference answers and improve accuracy. You can find more episodes humanloop.com/podcast

🚀 We’re excited to announce that Humanloop has been recognized as a Challenger in Gartner’s new Emerging Market Quadrant for Generative AI Technologies! Our mission is to enable teams to collaboratively build AI applications that are scalable and trustworthy. Companies like Gusto, Vanta and Duolingo depend on Humanloop’s unified platform for prompt management, observability, and evaluations to ship their AI features to market faster and reliably. The LLM/AI Engineering landscape is a rapidly evolving space and we love helping our customers navigate this new paradigm. Reach out to for questions or book a demo (link in comments). Like Gartner says: 80% of the engineering workforce will need to be upskilled in the next few years. Organizations should be investing in AI developer platforms that help them integrate AI into their solutions efficiently and at scale.

As we discussed on the High Agency podcast with Hamel H., Your AI app needs evals! This guide is a deep dive on how to build those evals in practice written by Peter Hayes who's helped many companies get AI applications into production.

At #devday this week OpenAI released prompt caching - which reduces input token costs by 50% and latency by 80% 🤯 Not too long ago, Anthropic released its version of prompt caching, which can save up to 90% on input tokens and reduce latency by 85% 🚀 This is a huge deal. Not only because they will make AI applications faster and cheaper, but because they require you to re-think how you engineer your prompts! Depending on which model provider you use, prompt caching varies in terms of cost, control and complexity. To understand the full picture, read our guide: https://lnkd.in/esMAj38h

Tree-of-Thought (ToT) prompting enables LLMs to think more like humans. Unlike linear prompting techniques, ToT evaluates multiple paths to answering a given question simultaneously before choosing the optimal approach, leading to a significant boost in reasoning capabilities. To learn about Tree-of-Thought (ToT) prompting, how it works and why it matters - check out latest explainer: https://lnkd.in/eyxcDUzC

8 RAG Architectures you need to know: 1. Simple RAG 2. Simple RAG with Memory 3. Branched RAG 4. HyDe (Hypothetical Document Embedding) 5. Adaptive RAG 6. Corrective RAG (CRAG) 7. Self-RAG 8. Agentic RAG Our latest blog will help you understand each architecture and explain the most popular use cases for each. Read here: https://lnkd.in/eYRFfdEP

If you’re a PM building with AI, this episode is pure gold. Raz lays out clear, actionable advice for navigating the challenges of building an AI feature. Raz Nussbaum is a Senior Product Manager for AI at Gong — the leading AI platform for revenue teams. He is an absolute legend when it comes to building and scaling AI products that genuinely deliver value. In this episode, he opens up about what it takes to build successful AI products in an era where things change at lightning speed. Learn: 1. How LLMs changed product development at Gong AI 2. Why it’s important to include Product Managers in the prompt development process 3. How testing pre and post-deployment is different 4. What are the new challenges in the face of Generative AI 5. What’s next for Gong AI Tune in: → YouTube https://lnkd.in/eRc3YWgR → Spotify https://lnkd.in/eRkkTnMw → Apple https://lnkd.in/eVvTmg5i Some key takeaways: 🔶 Stop Talking, Start Prompting: Raz emphasizes the need for PMs to directly interact with the models. Forget the theoretical—get hands-on with prompt engineering, QA, and iterate based on real-world feedback as quickly as possible. 🔶 The Competitive Landscape is Tougher than Ever: With barriers to entry lower than ever, Raz shares his strategy to move fast, adapt quickly, and stay ahead in an industry where everyone is racing to innovate. 🔶 Real Customer Data is Your Best Friend: Before you even have a UI, test with real customer data. Raz gives you the blueprint for involving customers from the earliest stages and iterating in real-time to build features they’ll actually love. If you love it (which I think you will!), please take a moment to rate and review. It helps us reach more AI builders like you! P.S. Don’t forget to check out our past episodes humanloop.com/podcast if you’re looking to get even more insights into the world of AI and product management.

How does a traditional machine learning team transition into generative AI? Dixa was asking themselves this question in late 2022, when they realized they could achieve a lot more with less by leveraging LLMs. Using Humanloop, Dixa applied their machine learning expertise to generative AI app development, enabling them to evaluate and track key LLM performance metrics Now Dixa releases AI products 3x faster and achieves 95-100% accuracy rates across all LLM features. Learn more : https://lnkd.in/euJPWXTt