davidprat.com

David Prat Robles

Engineering leader — AI Infrastructure & LLM Platforms

Barcelona · remote EU

Portrait of David Prat Robles

Open to senior engineering leadership roles

years in engineering
15+
cost savings architected
€40M+
teams led in parallel
4
junior engineers mentored
40+
senior engineers mentored
30+
technical articles published
21
video tutorials produced
9
professional certifications
17

01

About

Engineering leader with over 15 years of experience leading cross-functional teams to build AI-driven products. My expertise spans product strategy, cloud infrastructure, and execution across fintech, aviation, and large language models — bridging engineering, design and data science to drive outcomes that serve both user needs and business goals.

Enterprise AI — Multiverse Computing
Led cross-industry AI initiatives solving foundational problems for clients in healthcare, banking and energy, resulting in €20M+ in new revenue — and roughly 50% average budget savings for clients through quantum-inspired LLM pruning and healing techniques.
Aviation innovation — IAG
Directed the development of technical products that optimized aircraft maintenance workflows, generating over €40M in measurable cost savings.
Entrepreneurship — Datapta & Finlooker
Founded Datapta, a consulting company delivering cloud and data-platform engagements for AUDI AG, Antimetal and Nebula Strategy, and built Finlooker — a macroeconomic-intelligence platform — with a rotating team of 40+ interns and junior engineers. Relaunched Finlooker.io in 2026 on an AI-native stack.
Knowledge sharing
I regularly author tutorials on big data, AI and DevSecOps, and break down complex technical concepts on my YouTube channel.

Since July 2026 I have been building Finlooker.io independently — while completing my Executive MBA at EAE and exploring my next senior engineering leadership role.

Outside of work, I cultivate balance through therapeutic practices and support nonprofits that promote societal wellness and conscious living.

02

Experience

2026 — present

Accenture logo

Accenture

Enterprise Architect Manager

2026

Multiverse Computing logo

Multiverse Computing

Senior Engineering Manager — AI

  • Managed 4 teams (30+ ML engineers, 4 tech leads) building custom LLMs and SLMs for enterprise clients — around €20M in delivered revenue, plus open releases to the Hugging Face community.
  • Cut model size 50%+ at 2–3% benchmark cost with CompactifAI, saving clients roughly half their inference budget on average; extended model capabilities with tool calling, coding and multilingual support.
  • Coached tech leads and ran knowledge-transfer sessions for new team members, building a collaborative, high-performing culture across the teams.
  • Co-led the cross-department LLM pipeline from research to production API serving and on-device deployment.
  • Partnered with the CTPO on roadmap and OKRs, drove standardisation of data and metadata across departments, and owned GPU allocation across teams.

2025 — 2026

IAG logo

IAG

Senior Tech Lead · Interim Head of Infrastructure

  • Led DevSecOps architecture and ways of working across the MRO department, including rolling out AI-assisted code generation and review; guided new tech leads and hiring.
  • Led technical delivery of an aircraft-maintenance AI assistant (RAG, hierarchical agents, serverless ingestion) within a €40M cost-savings programme, with DevSecOps and IaC on GitHub Actions and Terraform.
  • Partnered with senior leadership on long-term product backlog planning.

03

Portfolio

Finlooker.io

Jun 2026 — present

Founder

A macroeconomic-intelligence platform: a bird's-eye view over financial markets, macroeconomics and liquidity worldwide, built for people who want the signal without the noise.

Next.js on Vercel and Render · GitHub Actions CI/CD · Storybook and Chromatic for visual regression · PostHog product analytics · built with Claude Code

Visit Finlooker.io

GPT-OSS 10B (Blackstar)

Apr 2026

Multiverse Computing

Compressed GPT-OSS 20B to half its size while preserving up to 98% accuracy, halving model utilisation costs; improved Spanish, coding and tool calling, and served the model over API with vLLM.

View on Hugging Face

Qwen3 0.3B (LittleLamb)

Feb — Mar 2026

Multiverse Computing

Compressed Qwen3 0.6B to half its size at up to 98% accuracy; enhanced Spanish and Android-friendly tool calling, shipped in the CompactifAI mobile app via llama.cpp and released on Hugging Face.

View on Hugging Face

04

Education

05

Certifications

06

Languages

07

Articles

I write about AI infrastructure and LLM internals.

All postsArticles on LinkedIn

08

Videos

Tutorials on Kubernetes, GitOps and cloud tooling on my YouTube channel.

YouTube — @davidpratr

09

Contact

The fastest way to reach me is a LinkedIn message.

Message me on LinkedIn