Mason McCabe — Senior Full-Stack Engineer, AI-assisted delivery

I built the AI pipeline a forty-developer practice ships on

Most organizations are still deciding which AI tools to trust. I built the pipeline mine prototypes against, co-wrote the standard behind it, and shipped the LLM products it was proved out on.

A design-to-code pipeline, adopted practice-wide
I built a Vercel v0 pipeline that lets product managers prototype directly against our project boilerplates, so a design arrives as code on the same stack the team will ship. Adopted across a forty-developer practice.
The org-wide standard for AI-assisted development
I evaluated Claude Code and Kiro against real project work, and I co-lead the standard the organization now develops against: which tools, where in the workflow they belong, and what still goes through human review before it ships.
LLM products built and hardened end to end
Front ends and hardened APIs for AWS Bedrock-powered chatbot and knowledge-base products — two proofs of concept delivered end to end as the sole developer.

40
developers
in the practice using the design-to-code pipeline I built
4
developers
led from architecture through production release
2
Bedrock builds
chatbot and knowledge base — front end to API, sole developer

Six months to deliver what the previous vendor needed eighteen for

Previous vendor, same scope

18months

The team I led

6months

Greenfield rebuild of a production sports-team housing platform — bed-level booking, document management, and event-triggered email automation on Next.js, NestJS, and PostgreSQL, replacing spreadsheet unit tracking and disconnected guest and vendor communication tools. I led four developers through architecture, review, and release. Live, with an active roadmap.

Summary

AWS-certified full-stack engineer working deep in the TypeScript ecosystem — Next.js, NestJS, React, and Node on PostgreSQL and AWS, with hands-on LLM integration through Bedrock. I led a four-developer greenfield rebuild of a production booking platform, delivering in six months what the prior vendor had not delivered in eighteen. I own code review and production releases on every engagement I lead.

Experience

Senior Software Developer / Tech Lead

Aug 2025 – Present

RevStar Consulting, Remote

Led a four-developer greenfield rebuild of a production sports-team housing booking platform, delivering full functionality in six months — a third of the time the previous vendor needed. Live now with an active feature roadmap.
Architected and built bed-level booking, document management, and event-triggered email automation on Next.js, NestJS, and PostgreSQL, replacing spreadsheet unit tracking and disconnected guest and vendor communication tools with a single platform.
Own code review and production deployments across the engagement. Stood up AWS infrastructure with Terraform and integrated SonarQube quality gates into CI/CD.
Built front ends and hardened APIs for AWS Bedrock-powered chatbot and knowledge-base products, delivering two proofs of concept end to end as the sole developer.
Technical lead in front of the client on multiple concurrent engagements: I lead discovery and requirements calls, produce estimates, scope features, manage the delivery board, and advise stakeholders on technical direction.
Built a Vercel v0 design-to-code pipeline adopted across a forty-developer practice, letting product managers prototype directly against project boilerplates. Co-lead the org-wide standard for AI-assisted development after evaluating Claude Code and Kiro.
Run the developer talent pipeline: technical interviews, hiring recommendations, and structured six-to-nine-week mentorships for incoming engineers.

Software Developer

Jan 2024 – Jul 2025

IClassPro Inc., Remote

Researched, designed, and helped build a microservice integrating the core platform with a third-party payment gateway, selecting architectural patterns for reliability and scale.
Owned the microservice's production release process, coordinating with QA to protect application stability across deployments.
Shipped features, fixes, and enhancements across a large-scale Angular, React, and Laravel class-management codebase on an agile team spanning QA, DevOps, and the Chief Architect.

Software Developer (Contract)

2023

RevStar Consulting, Remote

Six-week contract engagement delivering maintenance and ticket-based development for a client platform. Rehired in 2025 as tech lead.

Freelance Developer

May 2022 – Dec 2024

McCabe Software Solutions LLC, Remote

Built a desktop project-management application in Python, PyQt6, and SQLite with a conditional documentation checklist, automated document filing, and report generation that replaced manual weekly and monthly report assembly.

Selected work

DigSite Leads

2026

Solo build, from an informal brief

A family member described an excavation lead-capture problem in a few sentences. This is what came back: a public page where property owners scope a dig and drop a pin on the site itself, and a gated dashboard that works the resulting leads — filter, sort, status-track, export to CSV. The point of the exercise was how little specification it takes to get to something deployed and clickable.

The lead form with a pin dropped on an Austin map. A coordinate badge reads 30.27110, -97.74742, the address field has auto-filled to Nueces Street, and the submit button has turned orange and active. — Capture. Dropping the pin fills the address from the coordinates and releases the submit button — before it, the button reads “drop a pin on the map to continue”.

The admin dashboard: counts across the top reading 20 leads, 11 new, 4 contacted, 4 qualified, 1 rejected; a map of the Austin area with pins coloured by status; and a filterable table of leads showing project type, budget, timeline and status. — Pipeline. Every lead is a pin coloured by status, so density shows where the work is. Leads shown are seeded demo records.

Submit stays disabled until a pin is dropped, so every lead arrives with real coordinates instead of an address someone has to interpret. Nominatim reverse-geocodes the pin into the address field rather than the other way round.
Form fields mirror how excavators actually qualify work — project type, scope, timeline, budget — so a lead is triaged before anyone picks up the phone.
Deployed against a pooled Neon connection string. A direct one exhausts Postgres connections under serverless cold starts, which is the kind of thing that only shows up once it is in front of someone.

Next.js 16, React 19, Prisma, PostgreSQL, Leaflet / OpenStreetMap, Tailwind, Zod, Vercel

Live demo SourceDashboard password: dig-demo-2026

AI market-signal digest

2026

Solo build

A scheduled bot that tracks how the consulting services market is selling and delivering AI, then posts a ranked three-to-five item digest to Slack once a week — sized for a five-minute segment in a standing meeting. Not a general AI feed: every item has to answer one relevance question, and most candidates do not.

A Slack message from AI News Bot headed “Market signal — Sep 9”, listing five numbered items on Certinia, Deloitte, Kyndryl, Genpact and KPMG, each ending in a linked publisher — diginomica or CIO Dive. — One Wednesday's post. Every item is a services firm selling or delivering AI rather than a model release or a funding round, which is the entire relevance test the ranking prompt enforces.

Collection and ranking are split across two cron modes, so a whole week's pool is ranked in a single LLM call rather than one per day.
The model scores; code cuts. Minimum score and item cap are enforced outside the model, so a thin week posts two items instead of padding to five.
If the model is unavailable it degrades to the highest-priority sources, capped per source and flagged as unranked. A Slack failure leaves the pool intact and marks nothing as seen, so the segment retries rather than vanishing.
Keyword matching runs on titles only — matching snippets as well was tried and produced mostly vendor-to-vendor false positives.

Python, GitHub Actions, Gemini API, Slack webhooks, feedparser

Source

How I work

Code review and production releases are mine
On every engagement I lead, I own code review and I own the production deploy. On the current one that means AWS infrastructure stood up with Terraform and SonarQube quality gates wired into CI/CD — so the bar is enforced by the pipeline, not by whoever happens to be looking.
Releases coordinate with QA, not around it
At IClassPro I owned the production release process for a payment-gateway microservice, coordinating every deployment with QA to protect the stability of a platform other teams were shipping into.
I run the engagement, not just the code
I lead discovery and requirements calls, produce the estimates, scope the features, manage the delivery board, and advise stakeholders on technical direction — across multiple concurrent engagements.
I grow the engineers who come after me
I run technical interviews and hiring recommendations, and structured six-to-nine-week mentorships for incoming engineers.

Skills

Core stack

TypeScript, Next.js, NestJS, React, Node.js, PostgreSQL, SQL, AWS

AI and LLM

AWS Bedrock, Bedrock knowledge bases, LLM integration, AI-assisted development (Claude Code, Kiro), Vercel v0 design-to-code pipelines

Infrastructure

Terraform, AWS Amplify, CI/CD, SonarQube, Vercel, REST API design, Microservices

Practice

Solution architecture, Code review, Production release ownership, Estimation and scoping, Technical leadership, Stakeholder advising, Agile / Scrum, Jira, Azure DevOps

Also worked in

JavaScript, Python, Express, Tailwind CSS, Material UI, Angular, Laravel, React Native, MongoDB, SQLite, HTML, CSS

Credentials

AWS Certified Developer – Associate
Amazon Web Services
Mar 2026
Full-Stack Web Development Bootcamp
DigitalCrafts
2022 – 2023
Associate of Engineering Management Technology
University of Arkansas Grantham
2021