~/projects/sherpaedth-london-2026 · 2026.10.02 → 2026.10.04
S.H.E.R.P.A
A casualty evacuation command screen for the front line, where drones have taken away the golden hour. It runs offline on a single laptop.
Fuses drone detections and soldier triage reports into one graph and recommends a TCCC triage category; the medic decides in one click
Re-plans every affected evacuation route in the same minute a drone closes a road, fully offline on one laptop
Built at the European Defense Tech Hackathon in London; I built the Replay (a 251-minute run in 0.8 s) and pitched it
Decision Support SystemsMulti-source Data FusionEdge & Offline AIComputer VisionSpeech & Voice AI
~/projects/javilifefounder · 2025.06 → 2026.02
JaviLife
An AI planner that turns a goal into what you do each day, by voice.
4.7★ · 719 users · paying customers
2,000-person waitlist from organic video
Shipped solo to the App Store, first mobile app
AI AgentsSpeech & Voice AIMobile Development
~/projects/protect-childlst-upm · 2024.11 → 2026.09
PROTECT-CHILD
Clinical data harmonisation and federated learning for an €8.09M Horizon Europe project.
−93% time mapping clinical concepts, with an LLM + RAG pipeline in production
€8.09M Horizon Europe project with Georgia Tech, HL7 Europe and four European hospitals
Federated algorithms written from the math up, identical to the centralised result
NLPRAGKnowledge GraphsFederated LearningHealth Data Interoperability
~/projects/esa-adb-agentindependent · 2026.07 → 2026.08
Trustworthy anomaly agent
Anomaly detection on real ESA spacecraft telemetry that an operator can actually act on.
0 false alarms across 976,182 normal windows, with 99% precision
Names the right channel in 120/120 cases annotated by ESA experts
Reproduces ESA's published baseline (F0.5 = 0.949) on a laptop, no GPU
Anomaly DetectionUncertainty QuantificationExplainable AIAI Evaluation & GuardrailsEdge & Offline AI
~/projects/divefy[ongoing]
Divefy
A dive-safety chat that never makes up a number.
Finds the right passage for 83/84 exam questions (hybrid search, contextual retrieval, reranking)
Passes 70 of 84 PADI exam questions with Sonnet 5, graded by an LLM judge
Haiku 4.5 lands one question behind Sonnet 5 (69 vs 70 of 84)
RAGAI Evaluation & GuardrailsUncertainty QuantificationEdge & Offline AI
~/projects/devonfounder · 2026.03 → 2026.05
Devon
Voice and follow-up agents for private dental clinics.
~/projects/manipulensindependent · 2026.05
Manipulens
Spots manipulation in social media content, phrase by phrase.
~/projects/vanguardfounder · 2026.07
Vanguard
A deep tech community in Madrid.
~/projects/decisionaiactuaupm-22 · 2025
DecisionAI
Interpretable multi-agent systems for hospital decisions (idea stage).
★ Top 10 of ~200 ideas
~/projects/speech-enhancementrobolabo · 2024.05 → 2024.08
Speech enhancement for a medical wearable
Recovered an abandoned speech enhancement prototype for a wearable device and compared it with two architectures of my own.
~/projects/coloexpressretech-hackathon-2 · 2026.02
Coloexpress
Traceable prioritisation of colonoscopy waiting lists.
★ Winner, Health vertical★ Selected for the Healthcare cohort★ Selected for the Innovation Forum
~/projects/bouncerapart-incident-sprint · 2026.09
Bouncer
Checks what an AI agent is about to do before it does it.
~/projects/hackspain-voice-agenthackspain-2026 · 2026.09
HackSpain voice agent
A voice receptionist that handles a clinic's appointment calls.
~/projects/nasa-space-apps[upcoming]
Astronaut health monitoring
A health monitoring platform for astronauts, at NASA Space Apps Barcelona.
~/projects/master-mlucm · 2024.09 → 2025.07
Master's ML projects
Three machine learning projects from my MSc.
