In software development, change is inevitable. In the early 2010s, software development and design shifted toward shared component libraries and reusable infrastructure to reduce development costs. Developers used shared tools and built multiple applications on the same infrastructure. The developer experience meant working in sync to deliver software applications that met customer requirements.
As artificial intelligence (AI) technology contributes more to software development, AI tools and AI-developed software benefit from the same developer experience and design principles. Modern AI development must refocus on design principles that provide clear specifications and direction. Reliable products rely on quality design. Turning sophisticated AI models into reliable products requires a developer experience grounded in design principles that define a common direction, standards, tools, and reusable infrastructure.
Maitrik Patel has encountered that same problem in several ways throughout his career. His work has moved from large-scale user experience and design systems to developer platforms, machine learning infrastructure, and AI systems. Maitrik currently works at Apple, where his work focuses on developer experience for Cloud AI. Before moving into AI infrastructure, he spent over a decade at companies including Sony, DocuSign, and Disney, developing design principles that helped product teams build efficient, reliable products.
Software System Design to AI Platform Experience
Long before developer experience became central to AI engineering, Maitrik worked on a similar class of problems at the intersection of software engineering and product design. At Sony, he worked on a global design framework supporting digital experiences across 110 countries and 30 languages. Building shared infrastructure at that scale gave product teams across markets a common engineering foundation without requiring each team to solve the same underlying problems independently.
Maitrik later brought that approach to DocuSign, where he helped develop Olive, the company’s design system for creating consistent product experiences across its applications. Design systems became increasingly important to large software organizations because they turned individual interface decisions into reusable engineering infrastructure.
Maitrik has faced the same need for a common developer experience, including shared infrastructure, tools, rules, and standards. As he moved into machine learning and AI infrastructure, the technology changed. Still, the core issues blocking reliable product development remain a lack of common, shared, and enforced design principles.
Developer Experience Changes With AI
In modern AI development, developers work with data pipelines, AI models, and tools that evaluate frameworks and code and enable automated deployment. Some teams use common tools while others each use their own. Standards may or may not be defined, so developers and AI agents may not be using the same ones. As developer environments and AI systems grow more complex, shared tools and standards become one of the strongest levers for building reliable products.
When the developer experience includes design principles that specify which tools, infrastructure, and standards to use, the code produced can be tested, debugged, and improved using the same shared principles and tools. In practice, those shared tools reach developers through a few layers: a CLI that brings the platform’s SDKs behind a single interface, IDE integrations that put the platform inside the environments where engineers already work, and reusable skills that let engineers apply proven practices with their choice of LLM harness and agentic workflow. Individual developers don’t need to find the best tool to debug AI code; they can use the same one. They spend more time developing and less time chasing tools.
Design principles ensure a common developer experience focused on quality code, rather than trying out and testing different tools, infrastructure options, and applicable standards. Developers focus on code, code quality, and meeting standards.
The newest challenge is agentic developer experience. AI systems are increasingly being designed not simply to generate an output, but to use tools, maintain context, and perform sequences of actions to answer a user’s query. An engineer debugging an AI agent may need to determine which data source it used, why it selected a particular tool, how it reacted as its context changed, and where a multi-step sequence began to fail. Testing and evaluation become part of the developer experience.
The Research Behind AI Agent Tooling
That challenge also connects Maitrik’s engineering interests with his research. He co-authored ASTRA-bench (arxiv.org/abs/2603.01357), a benchmark for evaluating tool-using AI agents under realistic personal-context conditions. The research evaluates agents across 2,413 scenarios involving time-evolving personal context, diverse tools, and complex user intents. Maitrik co-authored the benchmark with a twelve-member research team that included Samy Bengio. Since its release in March 2026, ASTRA-bench has been cited by more than a dozen subsequent studies, including independent benchmarks and surveys on tool-using AI agents. Its value lies in giving engineers a structured way to surface reliability failures that conventional, static evaluations can miss.
AI agents work under different conditions and use variable data. When queries include context changes, an agent may need to use multiple tools to find or generate an answer. During research, benchmarks capture conditions under which AI agents succeed and how well each performs in its environment.
The implications extend directly to developer tooling. If engineers want to build agents that operate reliably in real-world environments, they need evaluation systems that expose failures under those same conditions. Evaluation then becomes less of a final checkpoint after development and more closely integrated with building, testing, and debugging an AI system.
A Career Built Around Engineering Leverage
Across Maitrik’s career, one theme emerges: engineering leverage. At Sony and DocuSign, the challenge was creating reusable foundations that let many product teams build without independently solving the same interface and engineering problems.
In AI development, the abstraction layer is different. Still, organizations face a comparable question: which problems should individual engineers solve themselves, and which should the platform solve once and make available to everyone?
That distinction matters because the quality of internal engineering infrastructure can influence how quickly an organization learns. When developers can move more easily between an idea, an experiment, an evaluation, and a production system, they spend more effort solving the problem rather than reassembling the machinery around it.
Maitrik’s career, moving from global design systems to AI development infrastructure, reflects the continually evolving nature of software development as it merges with AI technology. The goal remains to produce reliable, high-quality products, but the path to get there keeps evolving. Today, he serves as a peer reviewer for leading AI venues, including NeurIPS, ICLR, CVPR, and ACL workshops, and mentors founders through Stanford GSB Ignite, Plug and Play, and Founder Institute. He also served as an ethics reviewer for NeurIPS 2026.
The Next Layer of AI Infrastructure
As access to capable AI models broadens, the infrastructure around them becomes increasingly consequential. Engineering organizations need to determine how developers experiment with models, evaluate behavior, diagnose failures, and ultimately turn them into reliable products. With agentic systems, those questions become even more complicated. Developers must reason about systems that interact with tools, change context, and operate across different external environments.
Developer experience moves closer to the center of the AI engineering stack. The next phase of AI development depends on creating more capable models and common infrastructure that lets engineers understand, evaluate, and test them effectively.
Maitrik’s career offers a useful lens into that transition. Design systems demonstrated that carefully constructed shared abstractions could multiply the effectiveness of large product organizations. AI is creating problems that benefit from applying the same design principles common in software system development. Reliable products are built using design principles that define standards, tools, and the infrastructure used by human and AI systems.
Rashan is a seasoned technology journalist and visionary leader serving as the Editor-in-Chief of DevX.com, a leading online publication focused on software development, programming languages, and emerging technologies. With his deep expertise in the tech industry and her passion for empowering developers, Rashan has transformed DevX.com into a vibrant hub of knowledge and innovation. Reach out to Rashan at [email protected]
