AI Production Systems: Transitioning from Pilots to Reliable Deployments
As the tech landscape evolves, many teams are shifting their focus from AI pilot projects to robust production systems. This transition is reshaping the discourse in AI engineering, especially as the first confirmed talks for QCon AI Boston, scheduled for June 1 to 2, hint at a concentration on the technical intricacies necessary for implementing AI in real-world conditions. While getting a demo to function is a notable achievement, the true challenge lies in building AI systems that are reliable, observable, explainable, and secure when deployed at scale.
- AI Production Systems: Transitioning from Pilots to Reliable Deployments
- Key AI Engineering Themes for 2026
- Context Engineering Over Prompting
- Agent Explainability
- Moving Beyond Basic RAG
- Bridging Offline and Live Performance
- Security and Governance
- The GenAI Platform Layer
- Noteworthy Speakers and Their Contributions
Curated by industry leaders like 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, Senior Engineering Manager at Zoox and author of MLOps with Ray, the program will confront a pivotal inquiry: what does it actually take to bring AI into production in a trustworthy manner?
Key AI Engineering Themes for 2026
The early agenda for QCon AI Boston showcases several prominent themes that are expected to dominate the conversation:
Context Engineering Over Prompting
Ricardo Ferreira, Lead, Developer Relations at Redis, advocates for a shift in perspective. He emphasizes that while prompts might shine in controlled demo scenarios, they often falter in real-world situations plagued by latency and limited context windows. By approaching AI as a systems design challenge rather than merely a prompt-writing one, teams can better prepare for practical applications.
Agent Explainability
The topic of Agent Explainability comes to the forefront with Hannes Hapke, Head of 575 Lab at Dataiku and ML/AI Google Developer Expert. He underlines the importance of understanding why an AI agent selects particular tools. When decisions lead to downstream failures, having insight into the decision-making process becomes essential—it’s not enough to rely solely on output logs.
More Read
Moving Beyond Basic RAG
Cassie Shum, Vice President of Ecosystem, Product Engineering at RelationalAI, will explore how knowledge graphs can elevate AI systems beyond mere retrieval capabilities. By integrating complex reasoning across entities, dependencies, and domain context, AI can operate at a much higher level, enabling richer insights and interactions.
Bridging Offline and Live Performance
Mallika Rao, Engineering Leader at Netflix, broadens the AI conversation to encompass the divide between offline performance evaluations and the chaotic nature of live user behavior. Through techniques such as inference and evaluations, she will address how to design systems that remain effective in the face of unpredictable real-world interactions.
Security and Governance
Advait Patel, Senior Site Reliability Engineer at Broadcom, takes a deep dive into the security aspect of AI by discussing Zero Trust Agent Systems. By creating systems that can pass rigorous audits while maintaining functionality, teams can ensure that AI systems are safely integrated into existing engineering and operational environments.
The GenAI Platform Layer
In a notable session, Siddharth Kodwani, Software Engineer in AI Infrastructure at DoorDash, and Swaroop Chitlur, Staff Engineer and Engineering Manager for Machine Learning Platform at DoorDash, will dissect the internal infrastructure necessary to support AI functionalities across teams. Their focus will be on elements such as retries, fallbacks, prompt versioning, and cost tracking, which are crucial for a well-rounded AI deployment.
Noteworthy Speakers and Their Contributions
Additional confirmed speakers include Francesca Lazzeri, Principal Group Director of Data and Applied AI Science at Microsoft, who will discuss trusted AI systems; Sudeep Das, Head of Machine Learning and Artificial Intelligence for New Business Verticals at DoorDash, who will focus on scaling consumer AI; and Niko Matsakis, Senior Principal Engineer at Amazon and lead designer of Rust, who will dive into AI agent development.
The discourse at QCon AI Boston signifies a pivotal shift: the focus is no longer merely on whether an AI model can produce impressive outputs. Instead, the conversation is turning towards the comprehensive systems that need to be established to ensure that these outputs are dependable and scalable under real-world production conditions. Addressing context management, reasoning, evaluation, observability, platform architecture, governance, and operational trust will be critical for teams navigating this complex landscape.
To stay updated with the latest insights from QCon AI Boston 2026, explore the agendas and discuss with peers at this pivotal event.
Inspired by: Source
