phospho (YC W24)’s Post

We never expected people to be talking about this... Using the phospho clustering, we analyzed the text generated by our AI assistant Tak. This automatically groups similar messages together, based on the user intention in the text. We discovered that someone was using our AI assistant Tak to generate dozens of travel guide descriptions! Definitely not what we expected when releasing the assistant. Do you know what the users of your LLM app are saying ?

We are thrilled to announce our integration with Langfuse (YC W23). This partnership brings together phospho's powerful text analysis capabilities with Langfuse's comprehensive LLM application development toolkit. In one click, you can now import your data from Langfuse into phospho, and keep it synced in real time. This collaboration empowers companies to build, monitor, and improve their LLM applications more effectively than ever before. Product and engineering team can now work together more seamlessly to create better AI solutions. Learn more and try it today at phospho[dot]ai

🌥 You can now deploy phospho in your own cloud! 🚢 We are releasing self-hosted deployment options on AWS, Azure, and GCP! 🔒 Data security and ownership are crucial for some use cases. We've listened to your concerns and are proud to offer a solution that puts you in control. In this configuration, no data will ever leave your infrastructure. 🛡 Key benefits of our new self-hosted option: - Complete data ownership - Enhanced security with your existing cloud infrastructure - Seamless integration with your current tech stack and authentication solutions - Customizable to meet specific compliance requirements 🔥 Whether you choose our cloud-based SaaS or the new self-hosted option, you'll still enjoy the same powerful features, stellar support and incredible iteration speed that phospho is known for. 😱 The best part? We can handle all of the setup and deployment for you! 📩 Looking to deploy phospho in your own cloud? Let's get in touch at [email protected] or shoot us a message.

💭 Automatic clustering of user messages is now available in phospho! 🔭 In just one click, phospho can now cluster user messages based on their intention and topics. 🔥 Companies of all sizes use phospho intent clustering to: - understand how people are using their product - detect new use cases for their product - identify unwanted activity 🚀 Try out for free the User Intention Clustering on your own data in the phospho platform. It is as simple as importing a CSV or Excel file!

The phospho team has been quoted in a research paper! A few months ago, during the San-Francisco Mistral AI hackathon, we built a new way to benchmark LLMs by having them compete in Street Fighter. We demonstrated that LLMs can play a video game like those typically solved by training NN policies through Reinforcement Learning. Two weeks ago, researchers from Germany explored using LLMs to tune particle accelerators and quoted our work (article attached to this post). They showcased that compared to the traditional RL strategies, using LLMs without specific training can produce encouraging results. This weekend, we are partners of the Mistral AI hackathon in Paris. What will you build? Congrats to Nicolas Oulianov, Pierre-Louis Biojout and Paul-Louis Venard from phospho and Stan Girard from Quivr (YC W24) for this achievement!

Welcome onboard Raphaël Slama!

phospho (YC W24) team was at the Elysée Palace yesterday, where President Emmanuel Macron gathered the French talents in AI. Proud to be part of such a vibrant community.

phospho (YC W24) live in Times Square! After San Francisco, our open-source monitoring platform for LLM apps is now conquering New York. We hope to follow the path of other French companies that have successfully broken into the Big Apple. Big shoutout to the Brex team for featuring us in this campaign.

🖼 Add vision to your LLM app, with just an API call. Does your LLM chatbot supports image inputs from users? Conversational interfaces are not just made of text. Today, users speak with screenshots, gifs, pictures… You can now enable your LLM app to see images with the phospho multimodal LLM API. Send an image and a prompt to the API, and in response get a description, a color, the text written, and more. Our flexible and efficient custom AI model quickly answers any query. All of these capabilities are also coming to the phospho platform! phospho multimodal LLM is generally available via API (link in the comments) or as on premises deployment.