Engineering Production AI: QCon AI Boston 2026 Focuses on the Practical Challenges of Deploying AI Systems

The AI landscape has evolved significantly in recent years, with organizations moving beyond experimental pilots toward deploying artificial intelligence systems in production environments. This transition marks a critical shift in focus from achieving impressive model outputs to building the surrounding infrastructure that makes AI dependable at scale. QCon AI Boston, scheduled for June 1-2, 2026, has released its early program, which directly addresses this evolution by concentrating on the engineering work required to create production-ready AI systems.

The conference, curated by industry experts including Eder Ignatowicz (Senior Principal Software Engineer and Architect at Red Hat AI), Meryem Arik (Co-Founder and CEO at Doubleword, previously TitanML), and Hien Luu (Sr. Engineering Manager at Zoox and author of MLOps with Ray), tackles a central question: what does it actually take to get AI into production in a way teams can trust?

Key Themes in Production AI Engineering

The early program reveals several recurring themes that reflect the current state of AI engineering challenges:

Context Engineering Over Prompting

Ricardo Ferreira, Lead of Developer Relations at Redis, will explore how prompts that ace demonstrations often fail under real-world constraints such as latency and limited context windows. His talk reframes AI as a systems design problem rather than merely a prompt-writing exercise. This shift acknowledges that successful production AI requires careful consideration of how context is managed, processed, and maintained throughout the system's lifecycle.

Agent Explainability

Hannes Hapke, Head of 575 Lab at Dataiku and Google Developer Expert for ML/AI, will address the critical need to understand why AI agents select specific tools. When tool calls are incorrect and failures propagate downstream, teams require visibility into the decision path itself—not just output logs. This focus on explainability represents a maturation in how organizations approach AI system design, moving beyond black-box models toward more transparent architectures.

Moving Beyond Basic RAG

Cassie Shum, Vice President of Ecosystem and Product Engineering at RelationalAI, will examine how knowledge graphs can elevate systems from simple retrieval to complex reasoning across entities, dependencies, and domain context. This evolution reflects the growing recognition that production AI systems need to understand relationships and context rather than merely retrieving relevant information.

Bridging Offline and Live Performance

Mallika Rao, Engineering Leader at Netflix, will broaden the scope beyond large language models to tackle the persistent gap between offline evaluation and messy real-world user behavior through inference, evals, and system design. This focus acknowledges that production environments introduce variables that cannot be fully replicated in controlled testing scenarios.

Security and Governance

Advait Patel, Senior Site Reliability Engineer at Broadcom, will focus on building Zero Trust Agent Systems that pass strict audits while remaining functional. This track reflects AI's integration into existing engineering and operational environments, where security and compliance cannot be afterthoughts but must be foundational elements of system design.

The GenAI Platform Layer

Siddharth Kodwani and Swaroop Chitlur from DoorDash will break down the internal infrastructure needed to support AI capabilities across teams, including retries, fallbacks, prompt versioning, and cost tracking. This platform-oriented approach recognizes that successful AI deployment requires robust infrastructure that abstracts complexity while providing essential observability and control mechanisms.

Production AI: A Systems Engineering Challenge

The confirmed speakers highlight a fundamental shift in how organizations approach AI. The question is no longer simply whether a model can produce an impressive output. Instead, the focus has shifted to whether teams can build the surrounding systems needed to make that capability dependable and scalable under production constraints.

This means managing context, reasoning, evaluation, observability, platform architecture, governance, and operational trust. Each of these areas represents a significant engineering challenge that requires specialized knowledge and careful consideration.

Additional confirmed speakers include Francesca Lazzeri from Microsoft on trusted AI systems, Sudeep Das from DoorDash on consumer AI at scale, and Niko Matsakis, the Rust lead designer from Amazon, who will discuss opening up AI agent development. This diverse lineup of practitioners ensures that the conference will provide practical insights from organizations that have successfully navigated the challenges of production AI deployment.

The evolution of AI from novelty to production system represents a maturation of the field, similar to how other technologies have progressed from experimental prototypes to reliable infrastructure. As more organizations invest in AI capabilities, the ability to engineer production-ready systems will become a critical differentiator.

For those interested in learning more about QCon AI Boston 2026, additional information is available at the official conference website. The event promises to provide valuable insights for engineers, architects, and technical leaders looking to build reliable, scalable AI systems in production environments.