I’m Frank. I’m a designer living in Berkeley, CA.

I grew up in beach-side Portugal, and prior to settling in the Bay, I lived in Berlin, and Los Angeles. I studied clothing technology, and my first job was designing a suit with wearable tech to protect you against catastrophes.

While I loved clothing tech, computers were all the more fascinating to me. I taught myself digital product and programming, and that's is what I’ve been doing for the last decade.

For fun, I like to write, sketch, and study.

I work at Superlist where, in response to the growing complexity of work software, we are re-imagining the checklist as a viable tool for knowledge work.

I joined to help with the fundraising pitch deck, and stayed to help shape the vision for the product, and to craft the website. Over the last year I found a soft spot for applying AI to the product, as it allows me to exercise my love for design while also satisfying my itch to build.  

Below are some projects I helmed at Superlist.

I first joined Superlist's founding team to help create the story for fundraising—both pre-seed, and seed stages adding up over to 13 million USD from Cherry and EQT.

Beyond fundraising, I frequently create decks to present ideas and designs, using them as a tool to help others see what I'm seeing more clearly.

With the goal to introduce the Superlist app, and encourage people to try it out, we created an experience that replicated the app itself on the site, kept content highly visual, and told a focused narrative about the product.

With the ultimate north star to build a list that makes itself, we fine-tuned an LLM to output the world's best productivity how-tos as simple, actionable lists.

We worked to make this Language Mode available directly inside Superlists, to allow anyone to seamlessly create action-oriented lists with tasks within the product.

The LLM's output was designed through the training data and since it outputs content blocks native to Superlist, the initial stage involved training it on how to format the output correctly.

As formatting performance improved, we expanded the dataset to activate the model toward short, actionable step-by-step instructions. Finally, we added some humor and emojis to align the response with the Superlist brand.

Common prompts at Superlist:

Our goal was to adapt a highly informative question-answer model like ChatGPT or Claude into a task-oriented, productivity-focused tool. The result, named Make, is a model designed to generate actionable checklists that put productivity first.

Introducing AI into Superlist posed the challenge of communicating AI without resorting to visual clichés. Striving for simplicity, we designed an icon that captures the essence of computers coming to life.

The icon was designed as a visual mnemonic symbolizing the convergence of human and machine—two circles merging to reveal a third. We strategically placed it throughout the app to indicate AI-powered features.

Finally, the same colors are used to briefly highlight text while waiting for the model's response.

With work scattered across email, Slack, meetings, and beyond, we need a checklist that brings these bits of work together. To accomplish this, we integrated with Gmail, Slack, and other tools to easily taskify information from other apps. One of the first integrations we worked on was a way to generate tasks from an email.

The purpose, content, objective, formatting, and recipients of an email can vary widely, ranging from a one-liner, to a 15-person discussion, and the all too common newsletter with plenty of 'interesting' links. A chain of prompts as well as programmatic work was put in place to turn this diverse input into a consistent output.

This system was split into a series of concurrent LLM tasks to find the sender, context, purpose, and actions in the email. Models were fine-tuned to handle each of these requests and assemble the final task, as well as summary. URLs were handled programmatically and hyperlinked in the final summary.

My resume can be downloaded here, and you can find me via email here. I appreciate you for stopping by.