- 01AI SaaS products
- 02Full-stack systems
- 03Workflow automation
- 04Product UI
- 05API-first architecture
A focused set of recent AI products and systems, each shipped to real users.
Dataset Report
Digital platform specializing in data management, exploration, and visualization, designed to transform complex information into structured datasets, interactive dashboards, and decision-oriented analysis tools.
Three practices that compound engineering, AI systems, and interface design.
Product engineering
End-to-end SaaS systems built for actual production load. Type-safe APIs, predictable data layers, and a frontend that survives real users without ceremony.
- Next.js
- Hono / tRPC
- Postgres
- Drizzle
- Stripe
- Edge / Workers
AI workflow systems
Image, text, and agent pipelines that respect latency, cost, and failure modes. The hard parts — queues, retries, observability — built in from day one.
- OpenAI
- Replicate
- Vercel AI SDK
- Inngest
- Queues
- Vector DBs
Interface design
Calm, opinionated product UI with restraint. Typography, hierarchy, and motion treated as engineering disciplines — not decoration applied at the end.
- Design systems
- Tailwind
- Radix / shadcn
- Framer Motion
- Figma
- Prototyping
Four moves, in order. Most of the work is removing things before adding them.
- — Step 01
Understand the business goal
Before any UI or schema. What does this product change for the people using it, and how do we know it worked?
- — Step 02
Design the smallest useful product
The shortest path between a real user and a real outcome. Everything else is deferred until the core is honest.
- — Step 03
Build with production architecture
Type-safe from edge to database. Observability, retries, and migrations as first-class — not bolted on under pressure.
- — Step 04
Refine until it feels effortless
The last 20% is where products stop feeling like demos. Latency, copy, motion, edge cases — sanded down until they disappear.
A small, durable toolchain I trust to take an idea all the way to production.
A boring stack, on purpose.
Stable defaults, opinionated where it matters, and replaceable where it doesn't. The result is a product that ships faster the second time and the tenth time.
- Next.jsframework
- TypeScriptlanguage
- React 19ui
- Tailwind CSSstyling
- Radix / shadcnprimitives
- Node.js / Honoruntime
- Postgresdatabase
- Drizzle ORMdata
- Cloudflareedge
- Verceldeploy
- Vercel AI SDKorchestration
- OpenAI / Anthropicmodels
- Replicateimage
- Inngestworkflows
- Dockerruntime
- Resendemail
- Stripepayments
- PostHog / Sentryobservability
About
I'm a developer who cares about both the system and the surface: the architecture users never see, and the interface they feel every second.
- Based
- Peru · UTC -5
- Practice
- AI-first SaaS products
- Years shipping
- 10+
- Availability
- Selected product builds
- Engagements
- Fractional · Build · Advisory
Have an AI product, SaaS idea, or workflow worth building?
Send a short brief. I'll help turn it into a focused, shippable product, usually within a couple of days, sometimes the same one.