Tomás R. Pita

I build AI and operations systems you can measure.

Senior software engineer and former CTO. I work with small teams that need AI, integrations and legacy systems to actually work.

Computer engineer since 2000, with earlier experience in Venezuela and professional work in Spain since 2019.

Book a callSend a message
Portrait of Tomás R. Pita
How a video ad moves through the pipeline I designed
  1. TranscribeWhisper
    1.2 s
  2. Retrieve precedentsMilvus
    0.4 s
  3. ClassifyLLM on vLLM
    2.9 s
  4. ValidateBusiness rules
    0.6 s

    Low confidence: goes to human review

Illustrative example: timings are not a real measurement. The real trace lives in Langfuse.

Work

E-commerce of equivalent perfumes · First technical hire

A middleware that replaced Sendcloud

Shipping labels went through Sendcloud, and its cost grew with every order. I designed and maintained an in-house middleware between the picking system, thirteen Shopify stores and five carriers.

It kept the Sendcloud input contract so the picking system did not change, and label requests are idempotent: if the sender repeats the call, no second shipment is created.

Gross savings of almost €20,000 a month compared with Sendcloud, at about 2,000 shipments a day.

Doze Dmas, a tech spin-off of InfoAdex · Tech Lead

Ad classification with generative AI

The manual process covered only 30-40% of digital ads and was far behind. I designed the AI microservice: transcription, precedent retrieval, classification and validation, with retries and a fallback when the database is down.

I measured the system against human reviewers and separated what the AI did not propose from what it proposed wrongly, so we knew where to improve.

Validated in a test environment. The production launch came after I left.

Technologies I work with

The ones I use or have used in production and in my own projects.

Generative AI

  • LangChain
  • LangGraph
  • RAG
  • Whisper ASR
  • vLLM
  • Milvus
  • Neo4j

Backend

  • Python
  • FastAPI
  • Flask
  • TypeScript
  • Node.js
  • NestJS
  • Express
  • Rust
  • Go

Frontend

  • React
  • Next.js
  • JavaScript
  • Tailwind CSS

Data and messaging

  • PostgreSQL
  • MongoDB
  • Redis
  • SQLite
  • Oracle
  • RabbitMQ

Cloud and DevOps

  • Azure (GPU VMs)
  • AWS
  • Google Cloud
  • Docker
  • Kubernetes
  • GitLab CI/CD
  • Nginx

Observability

  • Grafana
  • Prometheus
  • Loki
  • Tempo
  • Langfuse

E-commerce and logistics

  • Shopify
  • GLS · SEUR · UPS

How I can help

Three ways to work together, with fixed scope and price.

Generative AI in production
Pipelines, RAG, WhatsApp agents and evaluation. From prototype to something you can operate and measure.
Technical leadership for small teams
Architecture, documented decisions, hiring and mentoring. As a tech lead or part-time CTO.
Legacy system modernization
Move old corporate applications to a maintainable base without stopping operations.

EstancIA, what I build on my own

A WhatsApp AI assistant for vacation rental managers: it answers guests at any hour, checks real availability and alerts a person when needed.

Visit holaestancia.com

Let's talk

Book a call or tell me what you need in two lines. I reply within one working day.

Book a callLinkedIn

Or write to me here