arize

Agent skills for Arize — datasets, experiments, and traces via the ax CLI

Root cause analysis for AI agents. Detects agent loops, retry storms, and optimization opportunities in LangSmith, Langfuse, Arize Phoenix, and OpenTelemetry traces.

Plataforma civica anti-corrupcion para el Peru: 11 agentes de IA (Google ADK + Gemini 2.5 / Vertex AI) auditan contratos del Estado por ~1 sol. OCR con Document AI, RAG legal con Vertex AI Search, observabilidad con Arize.

Persona-based Agentic Testing Testing Platform

Next.js demo: Scalekit AgentKit tool executions and OpenAI LLM calls in one Arize AX trace — public-API wrapper, not a prototype patch.

NeuroScale 2.0 — Autonomous A2A Agent Layer for self-healing ML platforms. Built on KServe, ArgoCD, Kyverno. Google Cloud Rapid Agent Hackathon submission.

The audit layer for AI evaluation: a six-hat panel scores a submission against a rubric with evidence-grounded scores, then audits and self-corrects its own over-confident scores against a calibration prior — every step a traceable span in Arize AX. Gemini 3.1 Flash-Lite · Google ADK 2.0 · Arize AX. Trace it. Trust it.

Crash-test AI agents against a real ledger: a payment times out after it was processed, does the agent refund twice? Daytona + Amazon Bedrock + Arize. Interactive demo included (mock payments).

How I measure, improve and ship an AI agent: two eval loops, a release gate, and a method for LLM-as-a-Judge evaluators that a PM can steer.

Reproduce the one-neuron 'evil Qwen' steering on Qwen3.5-4B, then evaluate and guardrail it with Jev and Arize AX

Gemini-powered Green Coding Coach — reads real Arize Phoenix traces via MCP to detect AI token waste.

An intelligent virtual receptionist for Camping La Rosaleda, built with Google ADK (Agent Development Kit) and Vertex AI RAG Engine.

AI security case-file pipeline for Phoenix traces, Gemini analysis, and Cloud Run demos.

AWS AI BUILDER HACKATHON LAB: CRAFT WITH AI AND BUILD WITH AI - (Virtual Pet Store)

AI-powered student developer assistant with white-box reasoning, built with Gemini 2.5 Flash and Arize observability

Phoenix-compatible exporter: ragdrift's five drift scalars as OpenTelemetry span attributes for Arize alerting.

Rootin4 — Gemini agent that predicts who plays at the seat you bought for World Cup 2026. Built on Google ADK + Arize Phoenix MCP for the Google Cloud Rapid Agent Hackathon.

A Gemini agent that reads its own observability data via Phoenix MCP to avoid repeating past failures. Built for the Arize hackathon track.

Purpose-built AWS Strands + Bedrock agent over a real Bedrock Knowledge Base, instrumented and evaluated in Arize Phoenix: a custom OpenInference RETRIEVER span, three validated evals with span annotations + a dataset, and an automated feedback loop. The AWS + Arize 'better together' story, in code.