Skip to content

Contact me

LinkedInWhatsApp

Share this portfolio

Alan Staub Negro

Senior Full-Stack Engineer

AI-native · React / Next.js + .NET · platform

Close-up photo of Alan Staub Negro smiling

About

Senior full-stack engineer (~5 years) who ships end to end — React 19 / Next.js + TypeScript on the front, real .NET / PostgreSQL on the back, plus platform work (CI/CD, testing, observability). I work AI-native: Claude Code, LLM agents and MCP are part of how I ship, and I authored my team's shared AI-engineering conventions. At Debos, a smart-building SaaS, I owned frontend delivery and sustained it solo for ~8 months, and in my final stretch took over cross-department delivery coordination previously held by the CTO. Trilingual (Spanish native, French bilingual, English professional), remote-first, with a humanities background I bring to communication and product judgment.

Skills

AI & agentic dev

  • Claude Code
  • LLM agents (Claude, GPT)
  • MCP
  • AI harness
  • AI code review
  • AI conventions (CLAUDE.md)

Frontend

  • React 19
  • Next.js (App Router)
  • TypeScript
  • TanStack Query
  • React Hook Form
  • Zod
  • Zustand
  • Tailwind CSS
  • Radix UI

Backend & platform

  • .NET (C#)
  • EF Core
  • Node.js
  • PostgreSQL
  • REST APIs
  • CI/CD (GitHub Actions)
  • Vercel
  • Sentry
  • Amplitude
  • Playwright
  • Vitest

Also

  • Go
  • Svelte / SvelteKit
  • React Native / Expo
  • Python / Flask
  • Figma (UX/UI)

Languages

  • Spanish (native)
  • French (bilingual)
  • English (professional)

Experience

  • Debos logo

    Debos

    Smart-building SaaS · Full-Stack Engineer

    Employee · 10/24 - 08/26

    Stack

    • React 19
    • Next.js
    • TypeScript
    • .NET
    • PostgreSQL
    • Playwright
    • Sentry
    • Claude Code / MCP

    What I did

    • Owned frontend delivery; sustained it solo for ~8 months (3 large features to production, zero critical reverts)
    • Took over the tech department's delivery coordination — deadlines, unblocking, comms — a function the CTO handed over; the team's delivery benchmark
    • Built 6+ product modules end to end + a full-stack device search (React UI + .NET endpoint + PostgreSQL trigram indexes)
    • Cross-cutting architectural refactors adopted as team conventions; built the design system and type-safe i18n across the app
    • Built the end-to-end test suite (Playwright) from a dormant setup; added feature flags (Amplitude) + observability (Sentry)

    Highlights

    • Authored the team's AI-engineering conventions (CLAUDE.md + versioned code-review skills)
    • AI-augmented workflow: Claude Code, LLM agents, MCP
  • Nivimu wordmark

    Nivimu

    HR PWA startup · Frontend Developer

    Employee · 04/23 - 04/24

    Stack

    • React
    • TypeScript
    • Ant Design
    • Emotion
    • Redux
    • Redux Saga
    • TanStack Query
    • Bit
    • Mockoon
    • Sentry

    What I did

    • Owned features end to end with high individual responsibility — code review, branch self-approval, UX QA
    • Led the JavaScript → TypeScript migration across a large-scale HR PWA
    • Built an internal component library, design system and micro-frontends (Bit)
    • Handled time-zone-correct scheduling across the product
  • Memorízame logo — an M monogram

    Memorízame

    Spaced-repetition study app · Full-stack · design

    Freelancer · 12/22 - 04/23

    Stack

    • Svelte
    • TypeScript
    • Supabase
    • PostgreSQL
    • Figma
    • Miro

    What I did

    • Sole front-end developer; designed the product end to end — brand and logo, design system, UI/UX and layout
    • Owned the Svelte front-end architecture; collaborated on the Supabase back-end (PostgreSQL)
    • Built the SuperMemo spaced-repetition study flow
  • integro writed in black with the o in blue

    Íntegro

    Startup · Full stack

    Freelancer · 02/22 - 04/23

    Sole developer and technical analyst for a startup — company site end to end, design system and branding.

    Qwik · TypeScript · Figma

  • Attlos logo

    Attlos

    Startup · Front End

    Intern · 12/21 - 07/22

    First professional TypeScript/React role — built product features in an Agile/Scrum team.

    TypeScript · React · Redux · Azure

Trials

Company technical tests under a deadline — how I approach an unfamiliar problem, and the verdict each one earned.

  • Debos — Live problem-solving interview

    October 2024

    Hired on the spot

    Two-part technical interview; the real test was a design problem — a search box hitting a backend with highly variable latency: how do you handle it? My answer, three layers: a minimum query length, a 200–500 ms debounce, and an AbortController cancelling the in-flight request so a slower earlier response can't overwrite a newer one — the classic typeahead race condition. Hired on the spot; joined as an external consultant, then moved in-house. I later shipped exactly this in production — a full-stack device search (React + .NET + PostgreSQL trigram indexes).

  • Nivimu — Code Test

    From 20 to 21 March 2023 · 1 day

    Hired

    Filterable, sortable user-data table with a live summary card that re-syncs to the top row as sorting and filtering change. React + TypeScript + Redux Toolkit + Ant Design — delivered in a single day.

  • Logitravel — PrimeIT Technical Proof

    From 1 to 4 July 2022 · 3 days

    Best technical candidate

    Text-list manager with add / select / delete and full undo–redo. React + Redux Toolkit + styled-components, a centralized design-token system and SOLID structure — built in a 3-day proof.

Artifacts

Self-directed and open source — the proof behind the labels, not the labels themselves. One makes an AI produce correct code; the other is a product I designed and built end to end.

Open source · MCP · evals

Gotcha

An AI-native engineering harness — and the catalog of subtle bugs that compile, pass review, and ship anyway.

Flagship — robust ticket search over ~500k rows: min-length, debounce and request cancellation on the client; a trigram scan that stays a BitmapOr, not a Seq Scan, on the server.

The gap that matters isn't using an LLM to type faster — it's engineering the system that makes it produce correct code in a domain full of traps. Gotcha is that system: a defect catalog, versioned review skills, an MCP server that hands those rules to an agent while it writes, and an eval suite that measures whether the AI trips each trap.

  • 16-defect catalog
  • MCP server · 5 agent tools
  • Evals 11/16 · recall & precision 11/11
  • .NET · PostgreSQL · Next.js / React 19

Design engineering · open source

Memorízame logo — a two-tone M in teal and fuchsia

Memorízame

A spaced-repetition study app I designed and built end to end — brand, design system and interface, grounded in the science of learning.

Live and interactive — three modes (study, review, test), a real SM-2 scheduler, trilingual UI in light and dark. Try it right here.

The product I did this design work in is private, so I rebuilt it as an open, installable demo you can actually use — seeded with public-domain content. Every surface is mine: the two-tone logo, a tokenised light/dark design system (WCAG-AA / APCA-checked colour, three typefaces chosen for reading science), a trilingual UI and a real SM-2 scheduler. Held to the same senior bar as the rest — a design-token contract test and an axe-core accessibility gate fail the build on any regression.

  • Design system · tokens
  • SM-2 · offline PWA
  • WCAG 2.2 AA · APCA colour
  • Trilingual · light/dark
  • SvelteKit 2 · Svelte 5