Software engineer based in Madrid, building ML systems, developer tools, and performance-sensitive products from first principles.
Incoming Software Engineer Intern at Vercel in San Francisco for Summer 2026.
- ML systems: tensor/autograd engines, training loops, language-model experiments
- Developer tools and infrastructure: Rust backends, typed SDKs, schema-driven systems
- Product engineering: end-to-end web/mobile products with a bias toward fast iteration and clear UX
- Built a browser-side semantic search engine for WhatsApp using Rust + Wasm + Next.js, with local embeddings and hybrid retrieval.
- Building BareTensor, a near-zero-dependency tensor/autograd library with a C++ backend and Python bindings.
- Earned an off-cycle Vercel internship after isolating a TurboPack + Web Workers + Wasm issue and discussing the reproduction directly with Tim Neutkens.
- Codeforces: Max Rating 1800 (Expert). (Top ~5% globally).
- Meta Hacker Cup: 3x Top 2,000 Global Rank (Round 2 Qualifier).
- ICPC SWERC: 2x Regional Finalist (Southwestern Europe).
- Ada Byron: 2x Spanish National Finalist.
- Spanish Olympiad in Informatics (OIE): National Finalist.
- Madrid Regional Olympiad: Silver Medalist.
- BareTensor: Near-zero-dependency tensor/autograd library with a C++ backend, Python bindings via nanobind, a small
bt.nnstack, and end-to-end language-model experiments from MLPs to small decoder-only Transformers. - ChatVault: Semantic search for WhatsApp that runs a quantized MiniLM/BERT model fully in-browser using Rust + Wasm.
- Vexi: Type-safe RAG database with a TypeScript SDK/CLI, schema-driven workflows, and a Rust API over LanceDB.
- rust-mlp: From-scratch MLP library in Rust with deterministic training and allocation-free inference.
- Cogniba: Brain-training product built with Next.js and Supabase, used by 2,000+ registered users.
