Digital innovation & product executive who still builds.
I work at the intersection of product, technology, customer experience, design, and business strategy — usually where there's an ambiguous problem and something new needs to exist.
I currently lead customer experience, digital product, and innovation initiatives at AAPC. I also remain hands-on with architecture, UX, prototyping, software development, and AI-enabled products.
I tend to work from idea → working product → scale.
That often means:
- identifying an unmet customer or business need
- rapidly designing and building a proof of concept
- validating whether the idea is worth pursuing
- establishing the architecture and product direction
- transitioning proven concepts to teams for production and scale
Two current products and an open-source tool where I explore how AI, product design, and modern software architecture come together.
An API-driven consumer platform spanning web, iOS, and Android focused on journaling, habits, activity, health data, and personal development.
I designed and built the platform end-to-end, including:
- product strategy and UX
- web and mobile applications
- backend and API architecture
- subscriptions and integrations
- Apple Health / Health Connect
- AI-powered conversational experiences
- recommendation systems
- agentic workflows
- automated testing and production delivery
Vitros is also where I experiment with how AI changes the way products are designed, engineered, tested, and operated.
nextgig-alpha.vercel.app · private alpha
An AI-powered job-search agent that continuously searches, evaluates, and ranks opportunities against a user's career profile, preferences, and constraints.
I designed and built the platform end-to-end, including:
- conversational career profiling and preference learning
- automated daily job discovery and verification
- probabilistic screening and rubric-based job scoring
- AI-generated application support grounded in résumé evidence
- interview practice and career coaching workflows
- scheduled search and weekly review agents
- multi-model routing with automatic provider failover
- eval-driven model selection based on quality, cost, and latency
- usage-based credits, rate limiting, transactional email, and SaaS infrastructure
Built with Next.js 16, React 19, TypeScript, Prisma/Postgres, Vercel AI SDK, Resend, Vitest, and Playwright.
NextGig is also an experiment in making AI systems more operational: not just generating answers, but running repeatable workflows, making ranked decisions, and improving those decisions through evals.
github.com/seanhudson/evalkeeper · open source · @crowandraven/evalkeeper on npm
An eval harness for the AI features in a TypeScript app. For each feature, it keeps an answer key people confirm, runs any model against it, and compares models on accuracy, serious mistakes, cost, and speed.
I extracted it from NextGig, where it chose the models for 20 AI features. It includes:
- answer keys that people own, with model drafts kept provisional until confirmed
- per-model tests with time and cost for every answer
- checks graded as serious or minor, with partial credit
- a model board with bootstrap comparisons against the live model
- suggested live and fallback models, each with its reason
- staleness tracking when prompts, schemas, cases, or graders change
- versioned model switches with one-click rollback
- a weekly drift check on every live model
- user feedback that turns into draft test cases
- probability and calibration scoring for decision models
Packages for the Vercel AI SDK and Prisma, with an example Next.js app that runs without API keys.
Evalkeeper is my answer to a simple question: when a provider changes a model or a cheaper one ships, how do you know, on evidence, which model each feature should run on?
Much of my current work explores:
- conversational product experiences
- AI agents and reusable tool/skill architectures
- retrieval and recommendation systems
- multi-provider AI architectures and fallbacks
- AI-assisted product development
- AI-assisted testing and QA
- rapid prototyping
- human-in-the-loop workflows
A few principles have followed me through product, technology, design, and leadership.
Curious people ask uncomfortable questions because that's how you find the real problem.
Add value deliberately. Make clear decisions, move with purpose, and adjust course when the evidence changes.
Documentation and process matter, but working software and fast iteration matter more.
Trust, communication, and collaboration build better products than process alone.
Do fewer things, better. Better products beat bigger products, and focus beats feature creep.
Show up ready to move the work forward. Momentum is a shared responsibility.
TypeScript · Next.js · React · React Native · Expo · Node.js · Postgres · APIs · Vercel
LLMs · RAG · Agents · AI SDKs · Vector Search · Automated Testing
The tools change. Shipping useful things is the point.
LinkedIn · Vitros · NextGig · Evalkeeper · Sean Hudson Studio · Crow and Raven