Building EdgeMind: Experimenting with LLM Inference on Android
Edge MLOps experiments: implementing Phi-3 mini on Android from scratch with NNAPI acceleration, custom tokenization, and KV caching
Hallöchen, I’m Ignacio
An MLOps & GenAI engineer based in Valencia, Spain.
I build production AI systems FROM SCRATCH.
Scroll to explore
What I Do
I'm an MLOps & GenAI engineer specializing in building production AI systems FROM SCRATCH. I built an enterprise AI automation system for a major European bank that handled code generation, project management, and meeting automation - delivered working POC in under a month solo.
I focus on LLM deployment, edge AI, and AI infrastructure that scales to production. My work includes on-device inference optimization, multi-agent systems, and framework-agnostic ML pipelines. I've been tinkering with hardware since age 12 (Arduino, electronics) and programming since age 15.
Currently exploring edge MLOps - running LLMs on mobile devices with custom tokenization and hardware acceleration. Recent project: Phi-3 inference on Android achieving 4 tokens/second with KV caching and NNAPI.
5+ years experience
5 languages fluent
Valencia, Spain
Featured Projects
EdgeMind
experimental
Experimental LLM inference system for Android exploring custom tokenization, KV caching, and hardware acceleration from scratch
Python
ONNX Runtime
Edge AI
LANZAT
completed
NASA Space Apps Challenge 2025 hackathon project combining satellite data and AI to predict climate vulnerability in Florida
Python
FastAPI
ML Pipeline
Recent Thoughts
AI Talk at Hackerspace Valencia
Intro session on generative AI with practical tools and examples.
Building Augen: AI Vision for Accessibility
Development insights and challenges from building an AI-powered vision assistant for accessibility
Specialties
Want to have a chat?
Always up for a conversation about MLOps, GenAI, or building production AI systems. Whether you're looking to collaborate on LLM deployment, edge AI infrastructure, or framework-agnostic ML solutions, I'd love to connect.
I'm particularly interested in projects involving on-device AI, multi-agent systems, and AI automation - especially in fintech and banking. Building since age 12 (hardware) and 15 (code), fluent in 5 languages, based in Valencia.