Portfolio v2026.11

Maximiliano Valentin López

AI Engineer

Technical Lead at Nolab

Mexico City · Remote

Spanish native · English C1

I build fast, secure web platforms, AI agents that run in production, and low-cost AI workflows that let a small team deliver like a large one.


Contact

/cv

work

Nolab

I rebuilt this marketplace from the ground up and brought AI into its day-to-day operations.

What the company gained

A platform rebuilt for speed and quality

delivered
A full rebuild of the marketplace, with the most care put where nobody looks: the data layer. A single, clean data model of the business now powers the public site, the admin panel and every integration.
result
Pages load fast, the platform is reliable and the design is better. Bug fixes take hours, not days, and new features take days, not weeks.
  • Next.js
  • React
  • PostgreSQL
  • Drizzle ORM
  • AWS
  • Terraform

More than a platform: how the company communicates

delivered
A communication system that is not a chat: the system's design carries the message. Each role has one responsibility and clear boundaries, every handoff happens inside the work itself, and every change is logged where the next person will look for it.
result
Communication improved and staff workflows lost their friction. Nothing gets lost between teams, and work keeps moving.
  • system design
  • single-responsibility roles
  • capability permissions
  • activity logs

Secure by design, from the cloud to the last role

delivered
Security as one coherent design, not a checklist. The edge layer filters what reaches the cloud, every service and every person gets only the access their role needs, and releases ship through an automated pipeline where no one holds permanent credentials.
result
Built so nothing is exposed by default and a failed release rolls back on its own. The product and the company stay protected without slowing anyone down.
  • Cloudflare
  • AWS
  • Terraform
  • GitHub Actions
  • RBAC
  • least-privilege IAM
  • OIDC
  • automatic rollback

Work principles

Data-driven development

I introduced product analytics and made it the rule. Tracked user events and session recordings drive UX and UI changes, not opinions.

UX first

Easy to navigate, with the next step always obvious. If a user has to stop and think, we treat it as a bug.

AI-driven development

Coding agents implement, test and address review comments while I plan and approve. Low-cost infrastructure and per-call cost tracking keep a small team affordable and able to scale.

Quality over quantity

Short cycles, finished by hand. AI builds most of the volume, so human time goes into detail, review, testing and talking with the business.

Earlier projects

Apunta Techoffline

Appointment booking SaaS for clinics: agenda, calendar, patients and subscriptions.

  • Next.js
  • Stripe
  • Meta API
  • Google API
Apunta Tech agenda view with a week selector and hourly slots

Viste y Ruedaoffline

Cycling apparel store with stock per size, cart, filters and orders.

  • Next.js
  • Supabase
  • PayPal
Viste y Rueda product grid of cycling jerseys with sizes and prices

next ❯ /skills

skills

Languages

core
TypeScript · JavaScript · Python

Agents and LLM systems

runtime
Vercel AI SDK 7 · tool-loop agent · streamed UIturns finish server-side after the client disconnects, persisted sessions
tools
Zod-typed tools · narrowed schemas · result size capsa value the model must not invent is absent from the schema; bad input returns as a result so the model corrects itself
control
capability permissions · human approval gateagent output is a change request; the agent never writes live data
models
OpenAI text · vision · image · transcriptionone model per purpose, set in the database; yes/no checks by token probability
documents
vision OCR · markdown conversion · content-hash cachetool-based search and read for long documents
observability
OpenTelemetry · PostHog LLM analyticscost and failures recorded per call
safety
SSRF guard · malware-scanned uploadsprompts built from public fields only

AI engineering workflow

agents
Claude Code · picustom skills, slash commands, subagents, hooks, MCP servers, TypeScript extensions
orchestration
parallel agents · git worktreesdriven from the terminal, one dev server per worktree
testing
in-house browser driveragents log in with one-time codes and screenshot every step
review
Kodus AI review · GitHub GraphQLeach comment triaged as real bug or misread; an agent resolves the threads
guardrails
docs as source of truthagents never run Terraform, edit migrations or merge

Web platform

frontend
Next.js 16 · React 19 · Tailwind 4 · shadcn · ZustandApp Router, Cache Components, partial prerendering, Server Actions, React Hook Form with Zod, next-intl
backend
Node.js · NestJS · Elysia · FastAPImodular monolith, hexagonal, ports and adapters around providers
data
PostgreSQL · Drizzle ORM · MongoDB · Redisbackward-compatible migrations, rate limiting and locks
auth
NextAuth v5 · Google OAuth · one-time codes
integrations
Zoho CRM · EasyBroker · Meta catalogsoutbox with retries
testing
Vitest · strict type checks · dead-code detection in CI

Cloud and delivery

AWS
ECS Fargate · RDS · S3 · Lambda · CloudFrontALB, ECR, SES, SSM, Secrets Manager, SAM, CloudWatch
IaC and edge
Terraform · Cloudflaremodules per environment
CI/CD
GitHub Actions · OIDCmigration and build rehearsal on a copy of production, rolling deploy with circuit breaker
analytics
PostHog · GA4 · GTMconsent gate per jurisdiction
data pipelines
Apache Airflow · Databricks · Docker · Azurebatch and stream

next ❯ /experience

experience

Technical Lead at Nolab

  • Rebuilt the marketplace and its data layer; fixes now ship in hours and features in days.
  • Built the AI operations layer: an agent for internal staff, document intake, and background jobs that write, translate, tag and audit property listings.
  • Designed the roles, permissions and activity logs that keep the company in sync without relying on chat.
  • Lead delivery with coding agents and own the AWS infrastructure, managed with Terraform.
  • Next.js
  • React
  • TypeScript
  • Vercel AI SDK
  • OpenAI
  • PostgreSQL
  • NestJS
  • AWS
  • Terraform
  • Cloudflare
  • GitHub Actions

Full Stack Engineer at Finvero

  • Orchestrated batch and stream data jobs with Apache Airflow and Databricks.
  • Designed machine learning data pipelines with concurrency and message queues.
  • Standardized internal and external APIs with OpenAPI (Swagger).
  • React
  • FastAPI
  • Azure
  • Docker
  • Tailwind CSS
  • Apache Airflow

Backend Engineer at Grupo Salinas

  • Built CLI tools that automate Docker container management and AWS deployments.
  • Built a dedicated microservice for one-time password (OTP) auth, with locks and rate limiting on Redis.
  • Documented the API ecosystem so other departments could integrate with it.
  • NestJS
  • TypeScript
  • AWS
  • Docker
  • Redis
  • MongoDB
  • Swagger

Frontend Engineer at LikeNuuk

  • Built mobile-first admin dashboards for omnichannel management.
  • Prototyped and shipped MVPs for on-demand e-commerce.
  • Next.js
  • React
  • TypeScript
  • Node.js

QA Tester at Tortoise

  • Python
  • Selenium
  • Qt5
  • Tkinter

Frontend Developer at Rojo Mate

  • JavaScript
  • WordPress
  • HTML
  • CSS

next ❯ /about

about

How I became an engineer

I started by making games. Game Maker projects and game jams in 2017 and 2018 taught me to finish something under a deadline. I skipped the traditional university path and learned by building: first websites for clients at Rojo Mate, then Python automation to test teleoperated delivery robots at Tortoise.

What I have worked on

Frontend for e-commerce products, backend microservices for a large corporate group, then fintech: banking and lending platforms and the data pipelines behind their machine learning. Each move took me one layer deeper into the stack, which is why today I can own a product end to end, from the database to deployment.

How I work now

At Nolab I lead the engineering of a real estate marketplace and most of my day goes to planning, reviewing and steering coding agents rather than typing code. I write the docs first, keep iterations small, and never let an agent decide what a person should approve. The same rule shapes the AI I build into the product: narrow what the model is allowed to do, and it stops making things up.

Outside work

Cycling, hiking, photography and anime.

Where I am going

I want to grow into a web architect role and keep learning and building with each new generation of AI. I am open to remote roles.

next ❯ /email to start a conversation

? for shortcuts · / for commandsEN · ES ·