I’m a generalist builder who works at any level, from strategy and research to systems and visuals to . I help turn complex opportunities into solutions balancing simplicity and power.
I lead product design for Stately.
Stately enables whole organizations, from engineering and design to support and legal, to turn ideas into executable, deployable diagrams that reflect how their product actually works.
Working towards our ambitious vision has required thinking systematically about deeply technical domains, making difficult tradeoffs through iterative experiments.
I've helped design all aspects of the product, from the foundational nouns and verbs to detailed pixels, while also helping build our designs as a software engineer.
This screen recording 👉 shows a few steps in a typical workflow from what we have actually shipped so far. The user is creating an AI agent that can play Tic-Tac-Toe against itself through an executable diagram. A new "paused" state is added with an action to report each pause. The code is immediately updated for export. Then they simulate the logic to test it.
To see more about Stately and the rationale behind some of our decisions , check out a casual demo of our in-progress work that my engineering partner Farzad and I shared. Also find more images of Stately in my UI design gallery.
I am currently exploring what is possible now that we can use LLMs to perform fuzzy computations.
This includes helping release our early alpha Stately Agent tool for building reliable AI agents with state machines and the actor model.
I've also started making small SwiftUI prototypes to try out more fluid interactions with AI than we currently get with text inputs.
For example, one way AI will become more naturally integrated into our work is by allowing movement through spaces of models.
This fun experiment uses a latent consistency model to rapidly cycle through possible prompt responses by rocking a fidgetable block displaying the image.
After generating an image from a text prompt, pushing the image block sideways steps slowly through possible images while pushing vertically leaps further in the model space. Finger velocity determines the distance from the current point. The change is based on movement, not just location, enabling you to pause to view images.
This is just a start of an exploration. For example, this prototype made it clear just how nice direct manipulation of the image could be, but also had problems like imprecise control and obscured visibility. Experiencing it gave me the idea to try letting horizontal movement step sideways in semantic space while letting vertical movement explore closer to the current semantic state.
I led a product called Teachably from 2018—2021.
Instructional designers had no way to create remote-first training and workshop programs in a single, adaptive deliverable. The most common practice was just writing a PDF script for trainers to follow. The delivery of this content would be disconnected from activity to activity, inconsistent across groups, and, as I also learned, even stressful for the trainers.
But what if an entire session, whether a 1 day workshop or a 10 week training, could be built as an integrated, sharable package of slides, articles, activities, videos, and games?
To help solve this, I built Teachably from early explorations to enterprise sales.
Earlier in my career I designed for large enterprise initiatives across departments at SYKES. These projects involved researching new solutions in a variety of domains, testing prototypes to build support and align stakeholders for significant process changes, and other such responsibilities.
More info available on request
Next-gen workforce management
SYKES is in an industry with thin margins and complex processes. I worked with the CTO and others to design a new platform for running contact centers that would reduce costs and increase performance while being easier for agents to get help. After the successful prototype it grew to become a key piece in the company strategy, supporting hundreds of millions of dollars in revenue.
Help when you need it
When stranded on the side of the road with a broken car, the last thing you want to do is wait on hold for the next available customer service representative. I helped Honda with a streamlined flow that integrated the journey across phone calls, text messages, and a web app. The proposal was implemented and grew into a prominent service offering by the company.
An always available personal coach
Precision Learning was an initiative to improve agent performance and job satisfaction by using AI techniques to identify skill gaps then automatically deliver "microtrainings" through Slack at the time of need in between calls. New content could be gradually introduced, while spaced repetition reinforced prior training.
This meant no need to schedule classes or take whole groups of agents off the phones while agents could still continuously develop their skills. For example, the "refund compliance score" increased from 77% to 96% on a pilot program while people only recieved training they needed.
Expert-led curated learning
Executives and experts were unable to globally disseminate information to SYKES employees, so the Chief Marketing Officer asked my team to build a video-based learning platform. OneLEAP became a widely used internal tool, eventually growing to run core programs like new hire training.
Now that you've skimmed my background as a product designer, check out some more quick samples of my UI work.