▶ Watch ↗14:38
Conference affiliation: Arize
King helps develop Phoenix as Arize AI’s head of open source; the retained interview explicitly establishes this contribution, rather than merely relying on his pinned repository. Phoenix is a self-hosted AI observability and evaluation platform combining OpenTelemetry-based tracing, versioned datasets, experiments, a prompt playground, and prompt management. This is an Arize team project, not a personally owned repository or evidence of sole authorship.
github.com · Source ↗King helps develop OpenInference through Arize’s open-source initiative. It supplies semantic conventions and instrumentation for tracing model invocations and surrounding retrieval, tool use, and application activity. It complements OpenTelemetry, a framework for collecting application telemetry, and can export to any compatible backend rather than requiring Arize’s platform. The evidence supports team participation and leadership, not sole creation or personal ownership.
github.com · Source ↗Authored by King on January 5, 2025. He describes moving into paid open-source engineering roughly two years earlier after a career building closed-source products. The essay explains his initial skepticism and the business rationale he developed around bottom-up adoption, faster community feedback, architectural transparency, public scrutiny, and self-hosting that addresses enterprise privacy concerns.
mikeldking.github.ioCo-authored by Xander Song and Mikyo King in March 2025. It explains why Phoenix’s team moved from keeping prompts exclusively in application code toward versioned prompt management for reproducible experiments. A prompt configuration includes model selection, tools, response format, and invocation parameters as well as template text. Provider-native formats and copying or syncing configurations into code preserve developer choice without requiring an observability service in the application’s runtime path.
arize.comCo-authored by Mikyo King and Elizabeth Hutton in May 2026. This product-direction essay proposes making traces, evaluations, annotations, feedback, and experiments accessible through APIs, CLIs, GraphQL, and agent-facing interfaces. Its intended verification loop is to inspect failures, make changes, rerun workloads, and compare evidence. It distinguishes human-directed collaboration from proactive agents and argues for delegation calibrated to risk. Its forward-looking Phoenix Intelligence vision should not be confused with the subsequent June beta introduction.
arize.comCo-authored by Mikyo King, Roger Yang, Nancy Chauhan, and Anthony Powell in June 2026. It introduces Phoenix Intelligence (PXI) in beta and describes the team’s implementation: page-aware context, a browser-based shell, an authenticated GraphQL query client, writable scratch space, documentation retrieval, and specialized tools. PXI helps debug traces, build evaluators, optimize prompts, and run experiments. Persistent changes require approval by default; recording its own model and tool activity is optional and requires administrator enablement and user opt-in. Co-authorship supports technical participation, not sole authorship of PXI.
arize.comAuthored by King in July 2026. This guide integrates evaluations with pytest, Vitest, and Jest while recording cases, scores, experiments, and debugging traces in Phoenix. Its support-bot example distinguishes refusal requirements enforced by assertions from helpfulness scores tracked over repeated runs. The TypeScript example additionally gates CI using aggregate helpfulness and latency criteria. Content-addressed, versioned examples let test fixtures remain the source of truth without duplicating datasets on every run.
arize.comMikyo King is head of open source at Arize AI, where he helps develop AI observability and evaluation tools, including Phoenix and OpenInference. His work helps developers inspect how AI applications behave across model calls, retrieval, and tools—and test whether a change actually improves the system.
King, also known as Mikel King, worked as a senior engineer at Apple before joining the founding team at Engine ML, where he built distributed training infrastructure for deep learning. He joined Arize’s engineering team in December 2020, drawn to its focus on machine-learning fairness and transparency. He graduated summa cum laude from the University of Southern California with a degree in computer engineering and computer science.
His move into company-backed open-source development came around 2023 and required rethinking how a software business should work. After a career building closed-source products, King initially questioned why a company with a commercial platform should invest in software people could use for free. He came to see open development as a practical business strategy: developers could try tools without a sales process, inspect architectural choices, and run the software inside their own infrastructure. A larger user community also shortened the team’s learning loop. Users could report problems, contribute code, and help shape what the team built next.
That approach shapes the projects he helps build. Arize Phoenix combines tracing, evaluation, versioned datasets, experiments, and prompt management in a self-hosted platform. Traces record the sequence of operations an application performed, giving developers a way to investigate failures beyond the final answer. OpenInference supplies conventions and instrumentation for capturing model calls and surrounding retrieval and tool activity. It works with any OpenTelemetry-compatible backend, so adopting its instrumentation does not require adopting Arize’s platform.
In May 2026, King and Elizabeth Hutton outlined a direction for Phoenix in which traces, evaluation results, feedback, and experiments become context that coding agents can query through APIs, command-line tools, and other programmatic interfaces. Their proposed improvement loop is concrete: inspect a failure, make a change, rerun the workload, and compare the resulting behavior. Human oversight remains part of the design; teams should delegate more when failure modes are understood and retain tighter control when the stakes are high.
Phoenix Intelligence (PXI), introduced in beta in June 2026, puts that direction into a product. King co-authored its technical introduction with Roger Yang, Nancy Chauhan, and Anthony Powell. The built-in agent uses Phoenix telemetry, documentation, prompts, and datasets to help debug traces, create evaluators, and run experiments. It receives the context of the page a user is viewing, reducing the need to copy information into a separate coding-agent session.
The team built PXI around a browser-based shell, a query client, a writable scratch space, and access to documentation before adding specialized tools. Those primitives let it investigate questions beyond a fixed menu of features. Persistent changes are staged for approval by default, and PXI’s own model calls and tool use can be traced when recording is enabled. The same infrastructure developers use to investigate their applications can therefore help them inspect—and improve—the agent doing the investigating.
▶ Watch ↗14:38
Loading more talks and highlights…