Concept

In the Spring of 2025, I completed a Cloud Computing course at Union Adventist University. This course explored AWS’s serverless services, and part of the curriculum included a semester-long project to demonstrate competency using these tools.

Rather than complete the provided tutorial project, I opted to build something more ambitious that could meet a real user need.

The Platelette project started as a concept for a recipe sharing and creation web app where users could post their own recipes and discover new ones to spark creativity. I love to cook, and I’m always exploring recipe sites to find inspiration from both professional and amateur chefs. I wanted Platelette to reflect that joy.

Design

From the beginning, I wanted the site to be responsive and organized with a grid-based layout using modular cards. The card idea is great for allowing users to quickly browse recipes, but also connects to my own childhood experience reading family heritage recipes on index cards in my mom’s recipe box.

For the color palette, I chose a gentle light theme for the content, contrasted with bold red in the header and primary elements to evoke the association of appetite. Headings use a secondary ‘chocolate’ color and stand out with larger font sizes.

Typography features a playful logo font paired with a clean, readable sans-serif for body text. Action buttons use a bright blue to stand out and align with user expectations, as blue is widely used for calls to action.

Solutions

I developed Platelette using Next.js, TypeScript, and Tailwind CSS. I chose Static Site Generation (SSG) for ultra-fast load times, great SEO, and compatibility with simple S3 hosting architecture.

One major challenge was the form-based input for recipes. Even though I designed the input to support editable servings and better searchability, it ended up being too time-consuming for users to complete.

Exploring AI with AWS Bedrock

Toward the end of the course, I began experimenting with AWS Bedrock, a new managed service that supports Foundation Models. As someone who loves exploring LLMs like ChatGPT, Gemini, and Claude, I was excited to see what a new model, AWS Nova, could do.

The result was a fast AI-powered recipe creation flow where all the user had to do was provide a title to get started.

Using Nova Lite and Titan Image Generator v2, I built a Lambda function triggered by a protected API endpoint. It receives the user’s prompt, authenticates with an access key, and responds with a JSON-formatted recipe and a generated image.

Multi-Modal Input Flexibility

By the end of the course, Platelette offered several recipe creation paths.

Users can modify AI-generated recipes, swap out photos with uploads, and customize as needed. This flexibility allows people to choose the level of control and effort that fits their individual preferences.

Reflection

I’m really happy with how intuitive and flexible the recipe creation experience turned out. Searchable tags, full-text search, and social features like liking and following creators made it feel dynamic and fun.

When people see the AI creation tool, the typical reaction is, “That’s actually really cool,” followed by, “Can I try?”

That said, there are things I’d change.

The name Platelette was a quick pick under deadline pressure. It wasn’t my first choice, but it was available as an affordable ‘.com’. Unfortunately, it confuses people. Even my professor had trouble typing it in class. It’s constantly autocorrected to “platelet,” and Google often redirects searches as well.

The original landing page functioned, but it lacked visual interest. After a visit to Williams Sonoma, I realized I had missed an opportunity: I hadn’t included a single plate in Platelette’s design. I revisited the landing page and added CSS-styled plates and Framer Motion animations.

The result? A much more engaging and animated experience. Future plans include adding a ‘Modify’ button to existing recipes to allow users to create variations of existing recipes on Platelette with AI, as well as Google Adsense to generate revenue to offset the minimal cost of using the AWS services that power the site.