Full Stack AI Systems Engineer

Lautaro Waltersdorf

Full-stack AI systems architect with four years building for the web. I architect parallel pipelines that ship advanced work in days, not months. Daily stack: orchestrated Python workers (Scrapy), live Node UI, and AI integration wired in through APIs. I work AI-native with Claude Code, the Claude CLI and minimax, and own problems end to end — scope to deploy.

Selected work

Things I've built

01

3D Car Visualizer — orchestrated end to end

cars-3d.lautarowaltersdorf.com

What it is

An immersive, real-time 3D car configurator that runs entirely in the browser. I orchestrated the full 3D visualization end to end — scene, frontend and audio generated around Sketchfab models — then shipped it with a cinematic camera intro, studio HDR lighting and physically-based materials. No plugin, no app, just a link.

What you can do

  • Drag to orbit the car in full 3D
  • Change the paint colour on the fly
  • Hit the gas to rev the engine with synced audio
  • Go fullscreen for the showroom view

Tech stack

  • Three.js
  • WebGL
  • GLTF / GLB
  • Tween.js
  • HDR / IBL
  • Web Audio
  • Vanilla JS

Key features

  • ACES filmic tone mapping with image-based studio lighting
  • Cinematic tweened camera intro on load
  • Mobile-aware rendering — adaptive pixel ratio and quality
  • Interactive paint, fullscreen mode and engine audio
  • Loads from a single static URL, no backend required

Hard problem solved

The engine wind-down is the part that usually sounds fake. Instead of synthesising a pitch-down, it plays a window cut from the tail of the real rev recording — anchored to the decay and cross-faded into a looping idle — so deceleration sounds like physics, not a toy. Audio also self-initialises muted on load so consent survives a fast load, and a single device check scales back blur, lighting and reflection size to keep phones smooth without a forked build.

Read the full engineering write-up on GitHub

02

Media Monitor — AI data pipeline

media-monitor.lautarowaltersdorf.com

What it is

A parallel AI data pipeline built from a real PR-agency brief: orchestrated Python workers collect press coverage in parallel underneath, an AI enrichment layer dedups it and runs sentiment analysis and AVE, and a live monitoring UI surfaces KPIs, a mentions feed, share of voice and a geographic map on top. Switch between clients, browse by month and export branded reports. The live demo runs on a seeded snapshot so it loads instantly; in production these views are fed by the orchestrated ingestion and enrichment pipeline described below.

What you can do

  • Switch clients and browse coverage month by month
  • Read the live mentions feed and filter by sentiment, outlet and tier
  • Track KPIs — mentions, AVE, reach and share of voice
  • Open a mention for full detail and review alerts
  • Generate an executive PDF report to export or email

Tech stack

  • Python
  • Scrapy
  • Playwright
  • Vanilla JS
  • ApexCharts
  • Leaflet
  • localStorage
  • PWA manifest

Key features

  • Live KPI cards with month-over-month deltas
  • Sentiment breakdown, trend lines and volume charts
  • Share of voice vs. competitors and a geographic mention map
  • Filterable mentions feed with per-mention detail sheets
  • Multi-client switcher, alerts view and PDF report export
  • Light/dark theme, responsive layout and PWA install

Hard problem solved

The real work is the ingestion engine behind the dashboard: a Scrapy crawler with AutoThrottle, retry/back-off and a per-domain circuit breaker, escalating to Playwright only for JS-rendered, consent- and paywall-gated outlets. The trick was making Playwright session-aware — a persistent browser context per domain solves the consent/paywall handshake once and reuses it — then feeding that hydrated HTML back into the same pipeline (dedup, sentiment, AVE) as the cheap HTTP path.

Read the full backend write-up on GitHub

Scraper Infrastructure

The multi-layer data pipeline that powers the work above — a UI monitoring panel on top, background Python workers (Scrapy on cron, with auto start/shutdown) doing parallel collection underneath, and an AI intelligence layer exposed through API endpoints. It's the orchestration pattern I reach for: collect in parallel, enrich with AI, run it in production.

  • Playwright
  • Python / Scrapy
  • Node.js
  • cron
  • Ubuntu server automation
  • REST API

Personal tools

AI tools I built for myself

Small, focused demos I built to learn the patterns — each one wires an LLM to a real interface and ships as a working app. Open source so others can fork them.

01

English Basics

AI English tutor for Spanish speakers

A focused grammar and vocabulary tutor that explains rules in Spanish, shows examples in English, and quizzes you until the pattern sticks.

  • Spanish-first explanations of English rules
  • Adaptive grammar and vocabulary drills
  • Conversational practice with AI feedback
View on GitHub

02

English Interview

AI mock interviews with grammar feedback

A spoken-English interview simulator. It asks the questions, listens to your answers, and grades grammar, vocabulary and structure line by line so you know what to fix.

  • Voice-driven mock interview flow
  • Per-sentence grammar corrections
  • Actionable feedback after each answer
View on GitHub

03

Interview Practice

AI technical interview simulator

A technical interview simulator for coding roles — picks a problem, takes your solution, and runs the kind of follow-up an interviewer would: edge cases, complexity, trade-offs.

  • Curated problem bank with hints
  • Code review and follow-up questions
  • Targeted feedback on correctness and style
View on GitHub

Tech stack

The tools I build with

AI systems

I build with AI in the loop every day and design the orchestration, not just the API call — agentic workflows, multi-step pipelines and LLM-driven tooling. Claude Code and the Claude CLI run my dev loop; minimax for fast iteration. It's how I ship a lot, fast.

  • Claude Code
  • Claude CLI
  • minimax
  • LLM orchestration
  • Agentic workflows
  • AI-native dev

Frontend

Responsive, accessible interfaces with semantic HTML, modern CSS and JavaScript — plus React, Redux and Next.js for richer single-page apps. I care about motion, type and the small details that make a page feel considered.

  • HTML
  • CSS
  • JavaScript
  • React
  • Redux
  • Next.js
  • Three.js

Backend

REST APIs, real-time features and relational data. I scope the data model, wire up the endpoints with Node, Express, PHP or Python — and reach for Supabase when a project wants managed Postgres and auth out of the box.

  • Node.js
  • Express
  • PHP
  • Python
  • REST APIs
  • WebSockets
  • PostgreSQL
  • Supabase
  • MySQL

Infrastructure & tooling

I deploy and run what I build — configuring and administering VPS servers on OpenLiteSpeed, automating browser checks with Playwright, and keeping a clean Git history and a command-line-first workflow throughout.

  • VPS
  • OpenLiteSpeed
  • Playwright
  • Git
  • GitHub
  • CLI

Design tools

Design isn't a separate phase for me. I move from Figma to code directly, and reach for Photoshop and After Effects when a project needs custom imagery or motion.

  • Figma
  • Photoshop
  • After Effects
  • UI / UX
  • Responsive design

Experience

Where I've shipped

  1. Founder & Full Stack Developer

    Apr 2026 – May 2026 · 2 months

    Web Factory Studio · Entrepreneurial exploration · LATAM market (remote)

    Launched Web Factory Studio as a two-month market exploration — a web and native app studio idea I took from concept to real client conversations, testing whether there was demand for high-quality, design-led product work in Latin America.

    • Validated the concept hands-on: built creative demos, ran outreach and held client meetings to test real market appetite.
    • Learned the market wasn't ready — budgets and interest in the region didn't support the kind of work I wanted to deliver.
    • Owned the whole thing end to end: scoping, pitching, demos and client relationships under real-world constraints.
    • Took the honest read and pivoted — channeling that initiative and business sense into finding the right team to build with.
    • HTML5
    • CSS3
    • JavaScript
    • React
    • Three.js
    • Node.js
    • Native apps
    • UI/UX
    • Figma
  2. Full Stack Developer — Freelance

    Feb 2024 – Nov 2024 · 10 months

    W Diseño Web · Own business · Argentina (remote)

    Built complete, custom web solutions for a range of businesses — owning both the user-facing interfaces and robust backend systems end to end.

    • Delivered bespoke sites and web apps from scratch, tailored to each client's needs.
    • Designed intuitive, attractive UI/UX with a focus on advanced CSS and responsive layouts.
    • Streamlined clients' internal management, production and sales using Node.js, Express and WebSockets.
    • Configured and administered VPS servers on OpenLiteSpeed for secure, performant deployments.
    • HTML5
    • CSS3
    • JavaScript
    • React
    • Redux
    • Node.js
    • Express
    • WebSockets
    • VPS / OpenLiteSpeed
    • Figma

Education

How I got here

Full Stack Web Developer

Henry — Intensive Bootcamp

6-month intensive program

Aeronautical Engineering

Universidad Tecnológica Nacional

5 years · incomplete

Technical Degree — Electrical & Electronic Engineering

Escuela Técnica Nº 28 "República Francesa"

6-year technical secondary program

About

A bit about me

I came to the web from an engineering background — electronics, then aeronautics — and that's the lens I still bring to it: I like understanding the whole system, not just the layer I'm working on. It's why I'm as happy reasoning about a backend, an AI pipeline or a deployment as I am sweating the type and motion on the front end.

These days that systems instinct goes into AI. I architect parallel pipelines that ship advanced work in days, not months — orchestrated Python workers (Scrapy), a live Node UI, and AI integration wired in through APIs. I work AI-native with Claude Code, the Claude CLI and minimax, thinking about the orchestration around a model, not just the API call. It lets me ship more, faster, without giving up the polish.

Running my own studio for a stretch taught me the rest. Pitching, scoping and shipping for real clients under real constraints made me someone who owns a problem end to end and cares about the result, not just the ticket. I work best with people who hold the same bar.

I'm based in El Bolsón, Argentina, and work remotely. Always learning, and looking for a team where I can do my best work over the long run.

Contact

Let's build something together.

Looking for someone who can build AI systems end to end and make them feel good to use? I'd love to hear what you're working on.