Mike Carbone

In the wake of the GPT phenomenon, I explored the possibility of blending LLM generation with design systems. After a series of explosive tweets and demos, I focused heavily on sales and market research to better understand the potential. Initial interest was booming from multiple Fortune 500 design system maintainers looking for a way to drive onboarding and adoption. Within weeks, we had a product we could sell to teams.

Out of the gate, we were met with questions regarding data security and compliance. Noise from the legal side of AI started to spread throughout companies, increasing the friction through our sales process. Finding an interested buyer was just the beginning; surviving the procurement process proved to be a much more formiddable challenge.

While we navigated the sales pipeline, we reached a peak of ten contractors; four engineers, two designers, three sales, and myself. Together we expanded the product to include an API platform and connected dashboard for users to manage their AI assistant connections.

The product was excellent, but unfortunately as time went on, we realized it would be a harder sell than we initially thought. Our AI products were a cherry on top of a larger sundae, and creating evangelists to help us through their procurement process was extremely difficult for a non-essential product.

We continue to hunt for product-market fit.

• Retrieval Augmented Generation
• Vector Databases
• Custom Applications on Large Language Models

With the development of Large Language Models, I saw the potential for a new network of connected personal assistants. We developed a customizable server framework and plugin architecture that could enable structured inter-node communication.

On top of individual Deco servers were plugins: application manifests that extended the capabilities of an individual's server. Each plugin created new API endpoints and allowed for safe cross-app communication.

The original goal of using retrieval augmented generation and inter-server LLM communication was never realized. By the time I was testing RAG and vectors, Hermae was born.

• Turborepo
• Open Source
• Solid Protocol
• Standards Development

I created this website to act as my personal 'city'. All of my writing, thoughts, pictures, links, and whatever else I want, expanding in any way I want it to. I wanted to make publishing as easy as possible and remove all dependencies on third parties. Simply put, I wanted to make it easy to use and long-lasting. My true, personal space.

An affirmation of independence. So far so good.

• Astro

Frustrated by Retool and other no-code admin interface builders, I wanted software that automatically built interfaces based on the shapes of my API responses. We desperately needed an admin tool that enabled us to bring in non-technical team members for managing the content within BarHop. This would include Create, Read, Update, Delete, with role-based access controls built on top.

Following those goals, Interweave was created over the summer of 2023. It supported dynamic input forms, generated tables, managed permissions, private and public interfaces, encrypted token management, and so much more. It worked beautifully for our team.

Interfaces were created by defining a data shape within your backend and sending it in a request to Interweave. The application would use this structure to build the interfaces needed, which enabled continuous integration and support for rapidly changing data models. No need to manually fidget with no-code tools each update.

A critical mistake I made with Interweave was designing it as a cloud application. This would've been a great opportunity to ship as a local-first app. If I were to rebuild this platform, I'd architect the interfaces to rely on the validation library Zod as an input, and sell it as local-first code you can append to your project. In hindsight, money is made in enterprise, and we were not equipped to handle the privacy considerations around proxying HTTP requests from private APIs.

I still use Interweave today to manage the content within my various API-driven applications, including this website.

• Next.js 13
• React Server Components
• Streaming
• Domain-Specific Languages

After moving to Hoboken, my brother and I decided to catalog all of the bars and create a fun app for our friends. The idea was great! People were interested, our social media was blowing up, and the potential to monetize seemed clear.

We started transitioning from my brother's proof of concept website to an iOS app powered by our own API. However, with a more fleshed out infrastructure, we saw the first flaw in our hypothesis.

As it turns out, staying on top of the weekend events and drink specials of 60-something bars was rather difficult. There was no great way to know when a bucket of beer was $7 and when an owner decided karaoke was happening that night. We went as far as creating a partner program for bar owners and managers to participate in (pretty much a way for them to text us with updates), but that revealed problem number two...

Most bar owners just didn't seem interested. Even though, on paper, this app would bring them more business and customers, they were not only uninterested, but almost... suspicious? Hostile? The reception was tough.

From this point, I was pretty discouraged. I had taken a back seat with drinking and going-out culture, and felt like I was entering a rut of endless feature development. In hindsight, I should've communicated this better with my brother. I ended up leaving the work and focusing more on Interweave.

BarHop lives on, to my surprise. We created a few believers and they continue to work with my brother on the experience.

• Postgres
• Prisma

We needed a platform where students could answer homework questions and receive a score for submission through Blackboard. I led a team of eight former students to design and develop this platform. I fostered design collaboration, UX critiques, testing, and development. As part of this new system, we also incorporated complex checks using AST tree matching; an improvement over the pre-existing string matching check, which led to many 'false incorrects'.

Together, we shipped the experience and the system was ready for use. Unforunately, another professor disapproved of the styles and wanted to make changes. In our perspective, the changes were unnecessary and not an improvement. The professor was adamant. To stand by the work of the students, we refused to adopt the changes, and the platform was never deployed.

• Delegation
• Stakeholder Management
• AST Application

As my first full-time role, I managed our frontend design system and frontend infrastructure. I worked on major projects like an organization-wide TypeScript migration, launched new designs of major components like our Button and Anchor, built tooling for dynamic pictograms and generative TypeScript interfaces, among many other developer experience improvements.

I spent a lot of time mentoring and helping other engineers navigate the trickier parts of frontend development, especially when it came to using the established best practices and component library. Newsela pushed me to be a more technical engineer and a leader in the frontend space.

After the founder & CEO left in January, the company was plauged by layoffs and low morale. Leadership was flipping every two weeks, I had no manager for months, my team was shrunk from 16 to 3 people, and it felt like we lost a lot of friends. I learned a lot about leadership and what it means to support those under you. And consequently, the effects of leadership failure.

I'm overall very thankful for this job — I think it really shaped my perspective on what I want from an employer in the future.

• Component Engineering
• Design System Management
• TypeScript
• Vite
• Storybook
• Emotion CSS
• Abstract Syntax Trees

In the winter of 2021, I was asked by my former program director if I could lead instruction for two courses of IDM-222. We covered many topics such as responsive web development, advanced HTML and CSS techniques, progressive enhancement, and so much more. This was a class I previously took as a student of Interactive Digitial Media at Drexel University.

This position opened my eyes up to how much I love teaching. Helping others is a great joy, especially around software. This is one of my proudest accomplishments.

• Teaching
• Mentorship

If you read the section below, you'll know that such an epic party needed software to organize it. We built Gatsby to manage the hundred or so guests that were interested in coming to Bonechella.

We created a public invitation website where people can register with their phone number, and after confirming, would be added to a waitlist. As event manager, I could accept or deny people from being added to the guest list. Once on the guest list, I could blast out texts to each guest coordinating the logistics of the party.

Of course, I turned this into a platform where anyone could create their own epic event. I learned how to design scalable data models for efficient reads, enabling quick operations at scale.

Working with Twilio as our provider of SMS sending was a disaster. I didn't realize that working with Twilio would mean navigating the complexities of SMS segmentation, 'A2P 10DLC Campaign Approval' and the fragility of automated messaging systems.

• Scalable Data Modeling
• Stripe API
• Twilio API

The pandemic sucked. It cut my senior year of college short and it isolated me from friends at a time where I should've been transitioning into full-time adult life.

As my act of defiance, I created a community of friends to do nothing and everything with. My people.

This ultimately manifested into hosting Bonechella, the post-pandemic pool party of the year.

We crushed about 300 hot dogs that night. It was epic.

After meeting Dan Mall my senior year at Drexel, we decided we wanted to work together. He shared with me his ideas of a design token manager for enterprise design systems, originating from his agency work with United Airlines, Marriot, Exxon and more. He also introduced me to Leslie Camacho. With Dan leading design and sales, myself development, and Leslie operations, I knew Arcade had a good shot at becoming a huge success.

We spent the next several months iterating on an MVP and working towards launch. We discovered how to work together, how to build, how not to build, and eventually stumbled to market. Our tech was great. Our operations were not.

After conducting demos for Spotify, T-Mobile, Pfizer, Mailchimp, and so many other big players, reception was... mixed. We certainly had not found product-market fit and it didn't feel like we were on a great path.

With a long list of wishlist features growing after each demo, I thought we needed funding to grow the development team. My co-founders were hesitant. Eventually, without funding, we lost steam and each of us had to take care of other business to feed ourselves.

A great adventure, a great disappointment, and so many lessons learned.

• Sales Development
• Demos
• Next.js
• Sanity CMS

My first company, serving as an umbrella organization for all of my ventures and client-facing work. To this day, my most profitable venture.

• Business Formation & Operation

Ten months of senior project. Six team members. The result of Senior project challenges students to produce professional-level media assets for a team based senior project in a simulated real-world production environment. It requires a project that demonstrates the integration of the academic and practical knowledge the student has acquired in the overall field as well as in one or more specializations. Teams should be able to not only defend how they are producing their projects but also why they are making them in the context of contemporary society and new media theory.

Together we produced a mobile app launched on the Apple App Store and Google Play — the first team in our program's history to do so. The result was incredible and we achieved our goals of creating a privacy-focused social network.

As a side note, a few months before launch I discovered network effects, and subsequently, the near-impossible difficulty of growing a social network focused on privacy. That killed any hopes of commercializing the product. Yikes!

• Research
• Async / Await
• Leading Development
• Network Effects
• Docusaurus

I teamed up with my brothers to build a new social network — anonymous group chats with your friends.

My brother Kevin led iOS development and my brother Nick led design. Over the course of a few months, we squeezed time out of our days to chip away on the app. Before full promotion, we pivoted to a YikYak clone so that we could have these group chats with people around us.

We launched the app and not much happened. Some users here and there. We ended up shelving the project.

• GeoJSON
• Socket.io

While at my first co-op, I had a lot of extra time to browse the depths of the internet. I would frequent ProductHunt and Twitter, where I would find interesting things each day.

Tired of losing links in my emails and DMs to myself, I decided to create a new project called CarbonCollective. This would be the home for my favorite links on the internet.

This project ended up expanding to a playground of new technologies for myself to play with. Authentication, APIs, the MERN stack, Google Cloud storage, and so much more.

CarbonCollective was alive and referenced by many for years until the Heroku free-tier shutdown.

• Scrolling pagination
• Google Cloud

My first job as a software engineer co-op! I was tasked with leading an accessibility audit of our frontend, and contributed to many different features in our mixed Ruby on Rails and Elixir Phoenix stack.

I made $27 an hour. Incredible money for college.

• Elixir Phoenix
• Docker
• Database Migrations
• Agile
• Frontend Accessibility

Bus Buddy is an interactive mobile web-app designed for parents with elementary school level students to help battle against the long morning school commute. With this system, students will no longer have to wait at their bus stop, unaware of their bus' arrival.

Bus Buddy is a hardware & software solution that uses an Internet of Things (IoT) approach to eliminate elementary school transportation uncertainties. The system works by using Raspberry PI technology with GPS and RFID sensors which allows students to scan their school IDs as they board their bus, allowing their parents to see where their bus is at any given morning or afternoon via the Bus Buddy app. Bus Buddy also sends parents SMS notifications when students have safely hopped on the bus, and when their bus is running late, a notification is sent with the updated ETA. A live map view with the Bus’ location is also available on the Bus Buddy app. Bus Buddy offers that peace of mind that parents are looking for while also alleviating frustration for students.

I was the leader for the project, making sure all teammates understood the goals, timeline, scope, and communication for the project. I also helped the design team with making sure the application had a consistent design, cohesive with the branding message the team was trying to accomplish. In addition, towards the second half of the project I led all of development, mapping out and implementing core features such as child status management, SMS notifications, and live-updating interface for the dashboard and map.

• React
• MongoDB
• Node
• Express
• RaspberryPi

In the Spring of 2017, I switched from a major in Animation & Visual Effects to a major in Interactive Digital Media. I was scared that this transition would set me back behind my new peers. To overcome this, I spent the summer teaching myself how to code so that I could get a leg up on my classmates. Naturally I gravitated towards building a tool to help me in school: an MLA citation generator.

I hated EasyBib and the existing big players of the time. Riddled with ads, slow, and cumbersome. I dedicated my summer to building a better user experience.

By the time classes started again in the fall, I was proficient in HTML, JS and CSS, and was far ahead of my peers. GoBibit was launched and today remains the most popular thing I've ever built, used by friends until their graduations years later.

• JavaScript
• HTML
• CSS
• Python