Book a callDownload CV (PDF)

Agustín Vivancos

Senior AI / Data Engineer — LLM Systems & Data Pipelines in Production · Python / FastAPI · AWS / GCP

17 years shipping webAI-first · 11 LLM agents in productionPython · FastAPI · data pipelinesAWS · GCP · Docker · K8s · CI/CD

Summary

Senior engineer (17 years) building AI-driven, scalable data solutions in production. I design and operate LLM systems and the data pipelines around them in Python / FastAPI on AWS / GCP — RAG, agent-proxy orchestration, streaming and post-call analytics — with full observability (cost, latency, traceability), Docker / Kubernetes, CI/CD and microservices. I run 11 LLM agents live in production. I work AI-first and spec-driven (TDD, contract testing, ADR per decision) and I like getting my hands dirty in the code. Obsessed with measurable efficiency: voice-LLM at $0.04/min, latency −35%, compute −30%, Meta Ads ROAS 4% → 9%. Based in Spain, available immediately for an on-site Madrid / EU contract that transitions to permanent.

Selected projects

pilis.appLIVELLM + DATA

in production

Production LLM agent API + data & analytics layer · Python / FastAPI · built end-to-end

  • Versioned public API (OpenAPI / Swagger) for chat, query, case-analysis, training-feedback and call-summarization — consumed by a Next.js / TS web and a React Native (Expo) app.
  • Data & analytics layer: post-call analysis pipelines, model management, internal/external dashboards — including Meta Ads performance feeding the CRM and reporting.
  • Contract testing + TDD over every critical flow; CI-gated on each push.

MigroLIVELEGAL-TECH

2022 — 2025

11 LLM agents · voice-LLM · data pipelines · web, mobile & API · live at migro.es

  • 11 LLM agents in production (MCP + LangChain): legal research, case mgmt, paid acquisition, voice analysis, compliance, document processing, governance.
  • Voice-LLM call-center (ElevenLabs → Telnyx) → structured CRM records and analytics; cost brought to $0.04/min by routing to open-source LLMs.
  • Meta Ads MCP agent operating paid acquisition (ROAS 4% → 9%, CPL −24%); clean API layering for −35% latency and −30% compute spend.

Selected experience

Full-Stack / AI & Data Engineer · IndependentNow

2023 — present

B2B via own LLC · AI-first, spec-driven + TDD end-to-end

  • Ship Python / FastAPI microservices and JS / TS clients on versioned API contracts with TDD; LLM agents and data pipelines in production (OpenAI · Anthropic · MCP · AWS / GCP).
  • Technical leadership: migrations, safe refactors under coverage, observability, ramp-up of junior teams.

Founder & Full-Stack Engineer · Migro

2022 — 2025

Legal-tech for immigration processes in Spain · Madrid · Remote

  • Owned product & architecture end-to-end — 11 LLM agents, voice-LLM and data pipelines (see Selected projects above).

Expansion & Driver-Acquisition Manager · Uber

2015 — 2018

Contractor · Lima, Peru · on-site · 3 years 5 months

  • Drove driver-acquisition strategy, onboarding flows and data-led growth for city launches across Peru, Colombia and Ecuador; cross-functional work across ops, marketing and expansion.
  • Lead-acquisition sharing agreement with a top advertising agency (2016) — ~$2M saved in acquisition spend.

Founder · HoyRed

2014 — 2020

Salamanca, Spain · tourism web network + hotel & apartments · full P&L for 6 years

  • Web properties driving qualified SEO / content traffic to offline assets; booking funnels, pricing, channel management.
  • Kept an active portfolio of technical clients throughout — shipping on Python and Next.js stacks.

Founder · Impulsa Consultores

2010 — 2013

First company, founded at 22 · digital consultancy

  • Led a team of 15 across engineering, design and accounts; enterprise clients BBVA · Banco Santander · Inditex + 90 SMEs.

CTO · enterbio

2011 — 2012

Organic brand moving to D2C · Madrid

  • Online store, order pipeline and operational integrations: catalog, fulfilment, billing.