Jake Dorscht
Builder
I focus on designing and developing AI-enabled products and the engineering workflows behind them. My work is grounded in practical, real-world software systems, with an emphasis on building trustworthy, maintainable solutions rather than showcasing only polished demos.
In my practice, I pay particular attention to the behavior of probabilistic systems in production, where unpredictability and failure modes are the rule rather than the exception. I’m especially interested in the developer experience of working with AI, including how we evaluate these systems, observe and recover from errors, and ensure that human judgment remains central to outcomes. The boundary between AI employed within a user-facing product and AI supporting the process of building software is a crucial distinction in my approach. Balancing constraints, implementing guardrails, and building observability infrastructure are central to how I ship reliable tools like LyndaAI.
Much of my ongoing work aims to create environments where failure is observable and recoverable, and where human decision-making remains at the forefront. AI serves as an assistant, not a replacement. This perspective informs how I approach the broader landscape of modern software development: viewing emerging tools and techniques as means to augment, not supplant, careful thinking and deliberate engineering.