Dominik Kundel is a product management and developer experience professional whose career has focused on developer platforms, cloud technologies, artificial intelligence, and software tooling. He is a developer at OpenAI, improving the experience of developers building applications on OpenAI's platform. ^[4]

Education

Kundel attended Jacobs University Bremen, where he earned his BSc in Computer Science. He later received his MBA from the University of California, Berkeley, Haas School of Business in 2024. ^[2]

Career

Kundel began his career in 2012 as an intern at SAP's Innovation Center in Potsdam, where he developed web-based user interfaces for applications built on the SAP HANA platform and contributed development guidelines for future projects. In 2013, he joined Microsoft as a Software Development Engineer in Test intern on the SharePoint team in Redmond, gaining experience in enterprise software development and testing. During his studies at Jacobs University Bremen, he also worked as a student web developer for the university's Campus Life department, served as a teaching assistant for Java programming and software engineering courses, and led the university's Computer Science Club, organizing technical events and launching the first student hackathon that later evolved into the independent jacobsHack! series. In 2014, he participated in Google Summer of Code with CodeCombat, improving mobile programming capabilities for the platform's code editor, while also founding the jacobsHack! hackathon series, which continued through 2017.

In late 2014, Kundel joined Microsoft in Dublin as a software engineer on the Marketplace team in the Office Extensibility division. He worked on the Office Store and related in-application experiences, with responsibilities that included user interface development, usability practices, and support for open-source software adoption within the Office organization. After leaving Microsoft in 2016, he joined Twilio in Berlin as a Developer Evangelist, supporting developer communities through technical education, conference presentations, documentation, and developer outreach.

Kundel relocated to the United States in 2019, becoming a Staff Developer Evangelist at Twilio, where he continued producing technical content, demonstrations, and educational resources for JavaScript developers. In 2020, he was promoted to Principal Developer Advocate, establishing a dedicated developer advocacy program that represented developer feedback within the company, creating processes to improve the developer experience across multiple product groups, and helping to launch initiatives that enabled early product testing and developer engagement at conferences and product previews.

Beginning in late 2021, Kundel transitioned into product management as Interim Principal Product Manager for Twilio's serverless platform before assuming the role permanently in 2022. He oversaw products including Functions, Assets, TwiML Bins, Twimlets, CodeExchange, and related developer tooling, guiding improvements to platform reliability and advancing the serverless offering to general availability. In 2023, he joined Twilio's Office of the CEO on a special projects team that evaluated strategic initiatives and supported executive decision-making. He was later appointed Director of Product Management for Emerging Technology and Innovation, where he led product and design efforts focused on artificial intelligence initiatives, including AI assistants, AI transparency concepts, executive technology strategy, and early-stage developer programs for experimental products.

In February 2025, Kundel joined OpenAI as part of the Developer Experience organization. His work focuses on improving the experience of developers building applications on OpenAI's platform by supporting developer tools, workflows, and AI platform adoption. ^{[1] [3]}

Presentations

Best Practices

At WAWTech in November 2025, Kundel delivered a comprehensive overview of best practices for developing intelligent agents. He clarified that agents are systems capable of independently solving tasks for users by combining models, instructions, tools, and a runtime or harness to facilitate execution and the delivery of outputs. Kundel showcased diverse examples, including Codex, a software engineering agent; ChatGPT's ability to perform complex tasks such as creating spreadsheets and browsing web data; and Atlas, an AI-enabled browser that automates tasks. He emphasized starting with a focused approach by defining a single task, choosing an appropriate model—preferably a state-of-the-art one like GPT-5.x—and gradually expanding capabilities through prompt engineering, harness selection, and tooling. He advised leveraging built-in tools or functions, and underlined the importance of observability, evaluation, and guardrails to ensure safety and performance. For scaling, Kundel suggested adopting multi-agent architectures, handling longer tasks with resilience, and managing context effectively through techniques like compaction. He also highlighted the potential of voice agents for enhanced user accessibility, recommending architectures such as chained or direct speech-to-speech models. Finally, he stressed continuous iteration, impact measurement, and alignment of agent development with specific user needs, rather than blindly chasing the latest models or frameworks, to create robust, scalable, and user-centric agents. ^[6]

Building Voice Agents

At AIEWF 2025 in June, Kundel presented an in-depth overview of building voice agents using OpenAI's tools, highlighting the release of the TypeScript SDK for OpenAI agents, which mirrors the functionalities available in Python. He explained that voice agents are systems that can perform tasks independently on behalf of users by combining language models, instructions, tools, and lifecycle management within a runtime environment. Kundel demonstrated how the SDK supports features such as handoffs, guard rails, streaming input/output, context management, human-in-the-loop support, and native voice capabilities, including interruption, WebRTC, and WebSocket communication. He elaborated on two architectures for voice agents: chained models that utilize speech-to-text and text-to-speech, and direct speech-to-eo models with lower latency and improved contextual understanding, despite challenges such as reusing existing capabilities and managing complex states. Through live demos, he illustrated how agents process voice input, execute tools (such as weather and refund services), handle interruptions, and provide detailed traces for debugging. Best practices emphasized starting with small, clear goals, implementing guardrails early, and leveraging generative qualities for personality and tone customization. Kundel also covered advanced features such as delegation, handoffs, human approval, and multiple-agent management, as well as considerations for evaluation, cost management, and policy enforcement in voice AI development. The session concluded with practical coding exercises and encouragement for participants to experiment, aiming to make voice interactions more accessible, natural, and efficient. ^[7]

AI Strategy

At AIEWF 2024 in December, Kundel shared insights into their proactive approach to integrating AI into customer engagement solutions over the past year. He explained that their team, comprising 16 members across engineering, product design, and go-to-market functions, operated as a self-contained unit focused on exploring emerging technologies like AI, but was not solely an AI team. The team learned that AI innovation could be categorized as sustaining, which improves existing products, and disruptive, which creates entirely new markets; current AI advancements, especially with chatbots and agents, were on the disruptive end but faced quality and regulatory challenges that prevented widespread enterprise adoption. Initially, their work centered on rapid prototyping of solutions such as AI personalization and perception engines, which struggled to gain market traction due to their niche, experimental nature. To address this, they adopted strategies like shipping early under a sub-brand called Tulio Alpha to gather rapid feedback without compromising their core quality standards. They emphasized team curiosity, flexibility, and transparency, recognizing the importance of sharing learnings internally and externally to foster innovation and build customer trust. Ultimately, their lessons highlighted the need for customer-centricity, adaptive roadmaps, and a willingness to fail fast, enabling them to navigate AI’s rapid evolution while maintaining their reputation and preparing for future disruptions. ^[8]