Decomposing the Agent Orchestration System: Lessons Learned

Why It Matters

Reliable agent orchestration cuts operational expenses and accelerates AI product deployment, making it a competitive differentiator for enterprises. Embedding ethical safeguards ensures broader adoption and regulatory compliance.

Key Takeaways

• Durable, self‑healing agents reduce operational costs.
• Debuggable infrastructure outweighs flashy feature releases.
• Flyte’s open‑source model drives scalable workflow orchestration.
• UnionML simplifies MLOps for microservice‑based models.
• Fairness, accountability, transparency shape future agent systems.

Pulse Analysis

The rise of autonomous agents has shifted the focus from isolated algorithmic breakthroughs to the robustness of the surrounding ecosystem. Organizations deploying large‑scale agents now confront challenges such as node failures, data drift, and unpredictable latency. Bantilan’s keynote highlighted that without a self‑healing backbone, even the most sophisticated models become liabilities, inflating maintenance budgets and eroding user trust. By treating orchestration as a first‑class concern, firms can transform agents from experimental prototypes into production‑grade services.

Union.ai’s contributions, particularly Flyte and UnionML, illustrate how open‑source tools can democratize resilient AI pipelines. Flyte provides a modular, fault‑tolerant workflow engine that abstracts away infrastructure quirks, while UnionML offers a standardized MLOps layer for microservice‑based models. Together they enable data scientists to focus on model innovation rather than operational minutiae. The emphasis on debuggability—through clear logging, tracing, and statistical testing via Pandera—further reduces mean time to resolution, a metric increasingly tied to revenue impact.

Beyond technical resilience, Bantilan stressed the imperative of fairness, accountability, and transparency in agent design. As regulators scrutinize automated decision‑making, embedding ethical guardrails early in the orchestration stack mitigates risk and builds stakeholder confidence. Companies that adopt these principles are better positioned to scale responsibly, attract investment, and navigate evolving compliance landscapes. In sum, the future of agent orchestration lies at the intersection of engineering durability and ethical stewardship, a combination that will define market leadership in the AI era.

Original Description

Niels Bantilan (Union) Keynote at the Coding Agents Conference at the Computer History Museum, March 3rd, 2026.

Abstract //

Building agents isn’t just coding—it’s surviving infrastructure, and Niels Bantilan knows the brutal truth: durable, self-healing, debuggable systems beat flashy features every time, because if your agents can’t handle failure, they’re just expensive paperweights.

Bio //

Niels is the Chief Machine Learning Engineer at Union.ai, and core maintainer of Flyte, an open source workflow orchestration tool, author of UnionML, an MLOps framework for machine learning microservices, and creator of Pandera, a statistical typing and data testing tool for scientific data containers. His mission is to help data science and machine learning practitioners be more productive.

He has a Masters in Public Health with a specialization in sociomedical science and public health informatics, and prior to that a background in developmental biology and immunology. His research interests include reinforcement learning, AutoML, creative machine learning, and fairness, accountability, and transparency in automated systems.