MSc AI & Cybersecurity · 3× IEEE/EMNLP published · Speaker at PyConEs & EuroSciPy · Based in Barcelona 🇪🇸
Research paper submitted to EMNLP 2026 — I architected the framework and authored 100% of the code.
A next-gen GraphRAG system that replaces expensive LLM-driven retrieval with pure mathematical graph navigation, achieving O(1) constant-time latency at any dataset scale.
| LiteRAG | Microsoft GraphRAG | ||
|---|---|---|---|
| 💰 Cost | €0.01 | €25.28 | 99.9% cheaper |
| ⚡ Latency | 1.42s | 142.04s | 100× faster |
| 🎯 Accuracy | 0.798 | 0.657 | +21% better |
Python PyTorch Neo4j NetworkX Sentence-Transformers
AI platform that autonomously transforms YouTube videos into viral vertical clips. Multi-modal pipeline fusing Whisper (transcription) → Gemini (hook detection) → PyAnnote + MediaPipe (speaker-aware face tracking) → FFmpeg (cinematic rendering).
50+ paying subscribers · 2,500+ clips generated · Stripe subscription revenue
React 19 TypeScript Python Whisper Gemini API MediaPipe Firebase Cloudflare R2 Stripe
Full-stack PaaS I lead that lets researchers deploy distributed workloads across AWS & IBM Cloud via a web IDE with Agentic AI (MCP) — no terminal, no DevOps. Natural-language cloud operations, automated CI/CD runtimes, real-time telemetry dashboards.
50+ active researchers · 80% faster setup · 99.98% SLA
React TypeScript Python AWS (Cognito, DynamoDB, CodeBuild) IBM Code Engine Kubernetes Docker
Open-source Python framework for transparent parallel execution across thousands of cloud functions.
| Contribution | Impact |
|---|---|
| HPC Engine — Singularity + RabbitMQ backend on MareNostrum 5 (BSC supercomputer) | Published IEEE 2024 |
| Kubernetes Work-Queue — RabbitMQ-backed pod pooling replacing one-pod-per-invocation | 7× cold-start improvement |
AWS EC2 Standalone — Dynamic VM provisioning with Create/Reuse/Consume lifecycle modes |
New compute backend from scratch |
Python AWS EC2/Lambda Kubernetes Singularity RabbitMQ Docker Dask
| Project | Language | Contribution |
|---|---|---|
| Ollama | Go | Low-level runtime optimizations reducing model loading latency |
| Google Gemini CLI | TypeScript | Tool orchestration layer for autonomous agentic workflows |
| Dask CloudProvider | Python | New IBM Code Engine backend outperforming commercial alternatives |
💻 Full Tech Stack
| AI / ML | Cloud & Infra | Languages | Tools |
|---|---|---|---|
| PyTorch · TensorFlow | AWS · GCP · IBM Cloud | Python · Go | React 19 · Vite |
| Neo4j · Gemini · Ollama | Kubernetes · Docker | TypeScript · Kotlin | Firebase · RabbitMQ |
| Whisper · MediaPipe | Serverless · CI/CD | Swift · Java | FFmpeg · Stripe · Dask |