Rachel Cantor

Design Pro was the product initiative that drove Hype's pivot from a news platform to a creator-economy product, work that enabled the company's acquisition. I led it end to end with a team of 4 engineers, wearing both product and engineering hats: deciding what to build, aligning product, design, back end, and marketing around one approach, and shipping it.

The decisions I'm proudest of were about scope, not code. I ran a value/effort analysis to find the version that delivered roughly 80% of the user value for about 40% of the engineering effort, which let us ship a month early. The clearest example: instead of building a complex block-level theming system, I used direct user feedback to justify shipping page-level theming first, cutting scope without cutting the value users actually cared about. Design Pro became the foundation on which the pivoted product was built, and the capability that made Hype an acquisition target.

RecipeCast is a micro-SaaS I designed, built, and shipped as the sole founder. It casts recipes to TVs and smart displays so you can cook without touching a phone screen mid-recipe. I own the whole stack: a React and TypeScript front end, a Node.js back end, Python services for recipe scraping and extraction, the Wasp framework tying it all together, and a Chrome extension for casting recipes from any site.

The core engineering problem was real-time state synchronization across 3 surfaces (a web app, a browser extension, and a Google Cast receiver on the TV), ensuring consistency as the user moves through a recipe. I built an Authoritative Receiver Pattern on top of the Google Cast SDK so the display device holds canonical state, with optimistic updates on the controllers for instant feedback. It has paying subscribers today.

TidyText is a web app I designed and built for cleaning up and reformatting messy text: stripping junk formatting, normalizing whitespace and quotes, and fixing artifacts that appear when text moves between apps. The processing runs client-side in the browser, so nothing the user pastes gets sent anywhere, which matters for sensitive content.

I am currently extending it into a cross-platform desktop app with Tauri, which is where the fully offline, installable version comes in. The project is an exercise in building a focused utility that does one thing well with a fast, frictionless UI, and in shipping the same core to both web and native targets.

A framework-agnostic nutrition-label web component built with Lit and TypeScript. It's form-associated through the native ElementInternals API, so the chosen serving count submits through a native HTML form with no extra wiring, the way a built-in input would. The same build runs in a plain HTML page or a React app with no wrapper package to maintain.

The engineering work that matters here is where the boundaries got drawn. The data layer lives in a Lit reactive controller, a plain class that owns fetching, loading, and error state, rather than sitting in the DOM as a fake service element. Nutrition data comes in only two ways, a src attribute or a facts property, so concerns like choosing a dataset stay in the consuming page instead of leaking into the reusable element. Accessibility is built into the structure: only the nutrient rows are a real table with proper headers, serving metadata is a definition list, and updates announce through an aria-atomic live region. It's the kind of self-contained, form-aware, accessible primitive a design system is actually built on.

A proof-of-concept high-frequency trading blotter built to explore the performance ceiling of data-dense financial UI. It renders around 500 simulated row updates per second in a React and AG Grid interface without dropping frames.

The key technique is async transaction batching. Instead of letting each update flow through React's reconciliation, I batch updates and apply them through AG Grid's transaction API, bypassing React's render cycle for the hot path while keeping the rest of the UI in React. I also built institutional-grade theming to match the density and visual conventions traders expect. The project is a focused study on where React's rendering model needs help with high-throughput data, and how to give it that help cleanly.