Ivan Nardini
@ivnardini
AI/ML Advocate x Vertex AI @googlecloud | Vertex AI dude | Open Models, Ray & Eval | Startup Mentor | AI Champion Innovator |
🚀 ADK v1.8.0 just dropped! This release is packed with new tools for enhanced agent development and deployment. Feature highlights: 🔹 Agent Card Builder class to generate A2A agent cards for ADK agents 🔹 A2A ASGI conversion to turn any ADK agent into an A2A ASGI deployable…
We're inviting a small group of Series A founders to a 2-day, in-person summit in SF with Google DeepMind & Google Cloud. What's in it for you? ✅ Build your AI roadmap 1:1 with our engineers. ✅ Get hands-on with the latest Gemini tools. ✅ Provide direct feedback on Google AI…
What's better than a fine-tuned Gemini model? A fine-tuned Gemini model that can process 200,000 requests at half the cost. Batch Predictions are now supported for your custom models! We just released the docs! Check it out. > Documentation: cloud.google.com/vertex-ai/gene… > Notebook:…
We've just launched the all-new Google Developer Program forums, designed from the ground up to create a more unified and collaborative space for everyone building with Google. Key features 🤖 AI-powered by Gemini for smarter search & post creation. 🤝 Connect with peers in new…
An agent's abilities also depend on context. The challenge is choosing what to include in the context. If done correctly, the agent succeeds; if not, it becomes forgetful and confused. Including the entire chat history would hit limits and results in high costs, slow…
spent the last few days just building with vertex ai and it's been awesome. went from teaching gemini how to see an image, to making imagen create one from scratch, and then building a full chatbot.
Open models just got an upgrade. Vertex AI’s Model Garden now includes DeepSeek R1, with full tooling and infra support. Check out @ivnardini's update here: cloud.google.com/blog/products/…
LLMs give you “intelligence”, but you still need tooling and infrastructure to ship applications.
Unlike rivals, Gemini is plugged directly into Google’s enterprise stack (Vertex AI), making it instantly scalable for business apps. Reality: The moat isn’t just technical—it’s distribution through the world’s largest infrastructure platform.
Evaluating complex, multi-agent AI systems is a critical challenge. The A2A team at Google Cloud has released a practical guide on evaluating an A2A (Agent2Agent) Reimbursement Agent using Vertex AI Evaluation services. The introductory blog covers everything from setting up…
