Philipp Schmid is a Staff Engineer at Google DeepMind, building AI Developer Experience and DevRel initiatives. He specializes in LLMs, RLHF, and making advanced AI accessible to developers worldwide.
204 articles from this blog
EvoCode-Bench is a multi-turn coding benchmark that tests AI agents on evolving tasks across 227 rounds, evaluating persistence and adaptability.
Learn how to use GitHub CLI inside Gemini API Managed Agents securely by injecting tokens via egress proxy without exposing them to the sandbox.
Guide to controlling an Android phone using Gemini 3.5 Flash's Computer Use feature with Python SDK and ADB.
Guide to using Google's Gemini Interactions API with JavaScript, covering text generation, streaming, multi-turn chat, and multimodal inputs.
Explains how Gemini Managed Agents work internally, including sandboxing, execution loops, and tool calling.
Guide to building and deploying AI agents using Gemini Managed Agents via the Interactions API, with code examples.
Explores four subagent patterns in AI agent systems for 2026, focusing on agent management and task delegation.
Guide to using Deep Research with the Gemini API, including setup, collaborative planning, and running research tasks.
Learn two proven patterns for correctly using MCP servers with AI agents to avoid context bloat, higher costs, and poor performance.
Tips for writing effective agent skills, covering structure, descriptions, and instructions for better AI performance.
Guide to using Google's Gemma 4 open models with the Gemini API and Google AI Studio, covering setup, text generation, and multimodal features.
Analysis of Kimi K2.5, Cursor Composer 2, and Chroma Context-1 reports on training agentic models with reinforcement learning.
Learn how to combine built-in tools like Google Search with custom function calling in the Gemini API for multi-step agent workflows.
A developer guide for using the Gemini Interactions API with the Nano Banana 2 image generation model to create personalized, search-grounded images.
Explores how autonomous AI agents (autoresearch) can optimize small language models by running hundreds of experiments overnight, improving performance without human intervention.
A guide to systematically evaluating and testing AI agent skills, covering success criteria, building an evaluation harness, and improving skill performance.
A guide to writing effective AGENTS.md files for AI coding agents, based on research data and best practices.
Explains the difference between an AI agent's inner loop (verifying work within a task) and outer loop (learning across tasks).
Guide to using multimodal function calling with Gemini 3 and the Interactions API to build AI agents that can process and analyze images.
A guide to using the Gemini Deep Research API for complex research tasks, including polling and streaming methods with code examples.
