Benjamin Matton at the Palace of Fine Arts in San Francisco.

I'm a software engineer living in San Francisco. I graduated from Wisconsin in 2024 and have spent the last year building VaBene — a group event planning marketplace — as the sole engineer on backend, web, and mobile. Before that, I shipped CI infrastructure at Nutanix and automated GUI testing at Geonexus. I like systems that are hard to get right: payment splitting, booking concurrency, multi-agent LLM workflows.

Currently building

VaBene

A group event planning marketplace. Solo-built across .NET, React, React Native, and Stripe Connect.

VaBene helps groups plan trips and events together — itinerary collaboration, native payment splitting through Stripe Connect, and a marketplace of services and experiences. I built the entire stack: ASP.NET Core 8 backend, PostgreSQL on Neon, React web, React Native + Expo mobile, 58+ Stripe Connect endpoints, real-time chat via SignalR, background jobs through Hangfire.

VaBene home screen showing upcoming events.
Upcoming events
VaBene Scottsdale Girls Trip itinerary, day-by-day with hike and spa activities.
Day-by-day planning
VaBene Napa Wine Weekend expenses with chip-in cost breakdown.
Stripe Connect payment splitting
VaBene group chat with a poll for booking a private chef.
Decide as a group

Projects

Animated terminal demo of the BeastChain blockchain mining a block.

BeastChain

A 3-node blockchain in Java, small enough to read in an afternoon.

ECDSA-signed UTXO transactions, proof-of-work mining at difficulty 4, gRPC peer gossip with broadcast-storm prevention. Refactored from a singleton-soup tutorial into a Node composition root — dropped integration test time from 5.6s to 140ms and surfaced a real bug where peer nodes weren't replaying UTXOs on accepted broadcast blocks.

  • Java 21
  • gRPC
  • BouncyCastle
  • Docker

github.com/benjaminematton/beastChain →

Streamlit dashboard scoring U.S. House donor impact across candidates.

oath_score — House Donor Impact Score

An ML model scoring U.S. House candidates 1–10 on donation impact.

Multi-quantile regression predicting race competitiveness on snapshot-date-disciplined inputs. Backtested on 2024: matches Cook Political Report's close-race accuracy using inputs available 53 days before Cook's final ratings (paired bootstrap Δ at T-60: −0.002, 95% CI [−0.34, +0.30]). 217 tests, end-to-end leakage integration test, Wayback Machine ingestion of contemporaneous Cook ratings to validate against back-fill.

  • Python
  • scikit-learn
  • Streamlit
  • FEC API

github.com/benjaminematton/election_predictions → live demo →

Telegram lead-card showing a qualifying Reddit post surfaced by the vabene-agents fleet.

vabene-agents

A small fleet of LLM agents scouting Reddit for VaBene's customer development.

Four openclaw skills running on a 24/7 Mac mini, scanning city subs every two hours and surfacing qualifying posts to Telegram for human review — nothing posts to Reddit automatically. Two-machine deployment discipline (dev laptop is the only commit-producer, host is a pure consumer). Killed the supply-side merchant-scout skill before shipping after discovering the data wasn't where the agent looked.

  • Claude API
  • openclaw
  • macOS launchd

github.com/benjaminematton/vabene-agents →

Experience

  1. Oct 2025–Present

    Founder & Full-Stack Engineer, VaBene Built three-platform product solo; 58+ Stripe Connect endpoints for multi-merchant payment splitting.

  2. Nov 2024–May 2025

    Software Engineer (Contract), Geonexus Cut UI test execution time 80% with a Jenkins automation framework.

  3. Summer 2023

    Software Engineering Intern, Nutanix Built a Spring Boot microservice auto-creating tickets for CI/CD failures across 700+ daily builds.

  4. May 2024

    B.S. Computer Science & Data Science, University of Wisconsin–Madison

Get in touch