Ashton Aschenbrener
Full-stack AI Engineer
I design and ship production retrieval systems, agentic workflows, and end-to-end features using Next.js, TypeScript, Python, and Postgres.
Currently open to early-stage roles where I can own features end-to-end — from UI to AI orchestration.
01 / Featured projects
Evidence-first projects.
Real systems, not slideware.
Portfolio-grade marketplace for discovering, sandbox-testing, evaluating, versioning, and exporting portable AI agent skills (SKILL.md).
What I built
- ›Marketplace discovery, filtering & skill cards UI
- ›In-browser sandbox runner with live traces & permission approvals
- ›Skill builder with validation, preview, test run & publish flow
- ›Version history with SKILL.md comparison & diffing
- ›Evaluation suites + regression indicators
- ›One-click install/export adapters for multiple AI tools & editors
Stack
Agentic job application system that scrapes candidate profiles, normalizes unstructured data, and scores role fit through traceable, evidence-based reasoning.
What I built
- ›Profile extraction pipeline
- ›Role matching workflow
- ›Multi-step agent loop
- ›Evidence-based scoring
Stack
Ontology-driven knowledge engine for financial concepts — combines semantic search with graph-based exploration and source-cited answers.
What I built
- ›Concept ontology with typed edges
- ›Hybrid retrieval (semantic + graph)
- ›Source-cited answer composer
- ›Coverage eval harness
Stack
Cross-provider AI evaluation platform that runs test suites, scores outputs against custom assertions, and surfaces regressions automatically.
What I built
- ›Full-stack eval platform
- ›Custom assertion rule builder
- ›Multi-provider runs (Gemini · Groq · OpenRouter)
- ›Regression detection + provider comparison
Stack
02 / Build log
What it looks like to ship.
A simulated build trace — clone, install, test, deploy. The commands and checkpoints that happen every time I ship something real.
03 / Demo
See the full experience.
An auto-playing walkthrough of the portfolio — hero, roles, build log, projects, and evidence. Each scene is a live view of what this site does.
Ashton Aschenbrener
I design and ship production retrieval systems, agent workflows, and evaluation infrastructure.
04 / Technical evidence
Skills and the projects that use them.
Every capability organized by area. Each item is something I've built and shipped — no hand-waving.
Frontend
App routing, responsive UI, accessible components
- ›Next.js app routing (App Router, server actions, route handlers)
- ›Responsive UI with Tailwind and accessible primitives
- ›Project dashboards with loading, empty, and error states
- ›Animated transitions with Framer Motion (intentional, not decorative)
Backend
API routes, Postgres schemas, validation at every boundary
- ›API routes and server actions
- ›Postgres schemas with foreign keys and event tables
- ›Server-side validation (zod-style) at every boundary
- ›Webhook intake with signature verification + replay
AI systems
RAG pipelines, agent loops, structured output schemas
- ›RAG pipelines (semantic + graph hybrid)
- ›Agent workflows with plan → tool → verify loops
- ›Tool orchestration with structured output schemas
- ›Eval harnesses with pass / partial / fail bookkeeping
Data
Knowledge graphs, vector stores, structured evidence logs
- ›Knowledge graphs and ontology design
- ›Vector + relational hybrid stores
- ›Per-step retrieval traces
- ›Structured evidence logs per match
Operations
Verification commands, failure handling, deployment notes
- ›Verification commands documented per project
- ›Failure handling per pipeline stage
- ›Deployment notes (Vercel, env, build)
- ›Logging and audit records preserved across actions
05 / Interested
Where I fit best.
Best fit: early-stage AI / product engineering. I own the whole stack end-to-end — UI, retrieval and agent layers, data model, deploys, evals.
What you get: projects that ship with eval harnesses, failure handling, audit logs, and a verification checklist — not a demo video and a prayer.