Alex Wick
I lead the Applied AI team at Constellation, an AI marketing platform for regulated industries. I joined in July 2025 as one of the team’s first two engineers.
Before that I studied computer science at Brown. I grew up in Burlington, Vermont, and now live in New York City.
At Constellation
Dealer pricing data
A product that tracks the lease and finance offers on the websites of all 18,000+ U.S. car dealerships. It reads offers from dealer-platform APIs where it can and uses LLM extraction everywhere else, with accuracy measured against a labeled benchmark.
Offer extraction agents
The first version of our offer extraction: a multi-agent LLM pipeline where specialized agents find, extract, and verify the offers on any dealership website.
Offer banners
An automated pipeline that launches offer banners across one automaker’s ~860 U.S. dealer websites, from picking which dealers get each banner to pushing it live. It replaced a manual process, and a request now goes live in four hours.
Billing
Constellation’s billing system, the source of truth for every client’s packages, products, and media budget. Other teams read from it through views, live spreadsheets, and APIs.
Internal platform
The Next.js, GraphQL, and AWS CDK app behind billing, pricing data, and operations. Its CI is built for coding agents: every pull request gets lint, type, and test gates and its own ephemeral AWS environment.
Earlier work
BETA Technologies
2024Software engineering intern at an electric aircraft maker. I built the factory floor’s nonconformance reporting app, cutting response time from 1–2 days to 30 minutes, and generated OpenAPI docs for 11,000+ ERP endpoints so teams could pull their own data for FAA certification.
Bluehouse Group
2022–24Developer intern at a digital agency in Burlington, through college. I rebuilt a local radio station’s website, which grew traffic 56% in three months, and designed the database schema for a vacation rental platform’s rewrite from legacy PHP to Laravel.
Projects
24cast.org
2023–24Brown students’ forecast of the 2024 U.S. elections. I recruited and led the five-person web team and built the front end, visualizations, and backend. It reached 30,000+ monthly users and was covered by The Brown Daily Herald, PR Newswire, and Bethesda Magazine. The site’s code is on GitHub.
Weenix
2024A Unix-like kernel in C for Brown’s operating systems course: processes and threads, device drivers, a virtual file system, an on-disk file system, and virtual memory with copy-on-write.
A small CNN trained from scratch (6.5M parameters) that matched a pretrained VGG16 at 75% validation accuracy on FaceForensics++ and Celeb-DF.