Suraj Vaghela

Suraj Vaghela · Seattle, WA · Founder, ALPHABYTE

I build products, and I ship them.

I'm an informatics student at the University of Washington and the founder of ALPHABYTE. I've put an app on the App Store and Google Play, built a marketplace that handles real money, and released a developer tool anyone can install.

Most of the code is private, so each project is written up here: what it does, why it exists, and where it stands.

About

I like the part where it reaches real people.

I'm a Seattle-based engineer studying informatics with a data science focus at the University of Washington, and the founder of ALPHABYTE.

What I enjoy is taking an idea all the way: sitting with the problem, designing how it should work, building it, and then getting it in front of people. A project doesn't feel finished to me until someone who isn't me is using it. That's why the work below is products rather than exercises: an app you can download today, a marketplace that moves real money, a tool other developers can install.

I care most about the parts users never see but always feel: that payments are correct, that private information stays private, that things keep working when something goes wrong. Getting those right is what separates a demo from a product.

Before ALPHABYTE I prototyped a Python ticket recommender at CloudifyAI Solutions, and I researched spiking neural networks as energy-efficient AI at Tech Mahindra's Makers Lab. I also work daily with AI coding tools, which is how I ended up building one of my own.

I'm looking for software engineering and machine learning internships or new-grad roles where I can own something end to end.

Education

Bachelor of Science in Informatics (Data Science)

University of Washington · Seattle, WA · Expected June 2027

  • GPA: 3.51/4.00
  • Dean’s List (2024, 2025, 2026)
  • iStartup Lab Micro-Grant (2026)

Coursework

Data Structures & Algorithms · Machine Learning · Artificial Intelligence · Research Methods

Certifications

  • Agents and Workflows OpenAI · 2026
  • Applied AI Foundations OpenAI · 2026
  • AI Fluency Framework & Foundations Anthropic · 2026
  • Generative AI Fundamentals Google Cloud

Verifiable on LinkedIn

Projects

Four things I've built.

Each one started as a problem worth solving. Here's what it does and where it stands, happy to go deeper on any of them.

Vouchlist

Live on iOS & Android ALPHABYTE

Local recommendations from people you actually know.

Vouchlist replaces the anonymous five-star review with a recommendation from someone in your circle. You find a plumber, a restaurant or a place to travel because a real person you know stood behind it. I'm the sole engineer: I designed and built the whole product and took it through both app store reviews.

The problem What came of it

Online reviews are anonymous and easy to fake, so it's hard to know who is actually behind a rating. Meanwhile the people whose taste you genuinely trust (friends, neighbours, family) aren't the ones writing them.

Shipped to the App Store and Google Play as sole engineer, reaching 76 users. Friends are matched without the backend holding readable contacts. Scout, an AI assistant (in beta with testers), answers plain-English requests like “affordable 4+ star restaurants in Australia” with friends’ vouches for home-service pros, restaurants, bars, and travel.

My role
Founder and sole engineer: design, product and engineering
Built with
Expo EAS · Supabase · Twilio Verify · Apple’s on-device LLM

ModelMarketplace

iStartup Lab Micro-Grant ALPHABYTE

A marketplace where developers sell the machine-learning models they have trained.

Training a good model takes real time and skill, but there's no obvious place to sell one. ModelMarketplace gives developers somewhere to list their work and gives buyers a trustworthy way to pay for it.

The problem What came of it

Developers who train genuinely useful models have nowhere natural to sell them, and buyers have no safe way to pay for something they can't inspect first.

Won a UW iStartup Lab Micro-Grant. Four Stripe payment options (purchase, lease, auction, seller subscriptions) and seller payouts via Stripe Connect.

My role
Founder: product and engineering
Built with
Next.js · Supabase · Stripe

unrot

Free and open source Developer tool

Catches out-of-date instructions before they mislead an AI coding assistant.

Teams now write instruction files to tell AI coding assistants how their project works. Then the project changes and the instructions don't. unrot reads those files, checks them against the actual code, and reports what no longer lines up.

The problem What came of it

An AI assistant following stale instructions confidently does the wrong thing, and nobody notices, because nothing ever checks whether those instructions are still true.

Published on npm. Measured accuracy by planting 356 known errors in real instruction files: it caught 354 (99.4%) and raised zero false alarms across 440 correct examples. Ran it on 69 popular open-source projects and reported findings upstream; fixes merged into Microsoft VS Code, PostHog, n8n, and OpenSRE.

My role
Creator and maintainer
Built with
TypeScript · npm

Kivo

Nearing App Store release Marketplace

Rent sports gear from people nearby, by the hour.

Kivo turns idle equipment into something rentable. Find gear near you on a map, book it, confirm the handover in person with a code, play, and pay only for the time you actually used. One account both rents and lends.

The problem What came of it

Trying a new sport means buying equipment you might use twice, while the gear other people already own sits unused in their garages.

A mobile app where users rent and lend gear, with map search, hourly booking, and handover codes. Bills renters only for time used via Stripe, with automated owner payouts; nearing App Store release.

My role
Founder: product and engineering
Built with
React Native · Stripe

Experience

Where the work happened.

  1. Mar 2026 – Present

    Founder & Software Engineer

    ALPHABYTE · Seattle, WA

    • Shipped Vouchlist, a friend-sourced local recommendations app, to the App Store and Google Play as sole engineer, reaching 76 users; built and submitted iOS and Android releases with Expo EAS.
    • Set up a CI/CD workflow with AI code review: a second agent reviews every GitHub pull request and either approves it or sends fixes back to the builder agent before merge.
    • Built Scout, an AI assistant (in beta with testers) that answers plain-English requests like “affordable 4+ star restaurants in Australia” with friends’ vouches for home-service pros, restaurants, bars, and travel, using Apple’s on-device LLM and returning only results the user has permission to see.
    • Protected user data with Supabase row-level security and SHA-256 contact hashing, so friends are matched without the backend holding readable contacts; added phone-OTP login (Twilio Verify).
    • Build every product by directing AI coding agents (Claude Code, Codex): own product design, feature planning, testing, and iteration while the agents write the code.
  2. Jun 2025 – Nov 2025

    Software Engineer Intern

    CloudifyAI Solutions, Inc. · Seattle, WA

    • Prototyped a Python ticket recommender on about 2,500 support tickets: NLTK text cleaning, vector embeddings, and top-5 retrieval of similar past resolutions by cosine similarity, improving on TF-IDF.
    • Built an n8n workflow that sent each incoming ticket to the Python recommender and emailed suggested resolutions back to the customer.
  3. Aug 2024 – Sep 2024

    Machine Learning Research Intern

    Tech Mahindra (Makers Lab) · Redmond, WA

    • Researched spiking neural networks (rate/temporal coding, STDP) as energy-efficient AI; presented my findings.

Toolkit

What I work with.

Languages

  • Python
  • Java
  • SQL
  • R

Platforms

  • Supabase (row-level security)
  • Stripe (payments)
  • Docker (test environments)

AI & developer tools

  • Claude Code
  • Codex
  • Cursor
  • n8n
  • Git & GitHub
  • Embeddings & semantic search

Contact

If any of this is close to what your team does, let's talk.

I'm open to software engineering and machine learning internships and new-grad roles. Most of the code above is private, but I'm glad to walk through any of it. Just ask.