I build AI products that solve real problems
From multilingual language learning platforms to intelligent agent systems — I design, build, and ship AI applications end to end.
Projects
AI applications built from problem to production.
Spralingua
AI language learning platform that delivers personalized lessons, real-time grammar correction, and natural conversation practice in 4 languages.
AI Voice Agent
Privacy-first voice assistant with 90%+ recognition accuracy, running entirely offline with zero cloud dependency.
AI Recruiting Assistant
Custom LLM assistant that automates candidate screening, interview preparation, and hiring decision support.
The Problem
Language learners plateau because generic apps lack personalized feedback and real conversational practice. Existing tools treat every learner the same, ignoring proficiency gaps and native language interference patterns.
The Approach
Built a full-stack platform using Claude's language understanding to generate adaptive lessons, provide contextual grammar corrections, and simulate natural conversations. Multi-agent architecture with specialized agents for vocabulary, grammar, and conversation.
The Result
A production application serving users across 4 languages (Spanish, German, Portuguese, English) with real-time speech-to-text, text-to-speech, and intelligent lesson progression. Sub-2-second conversation loop. Deployed on Railway with PostgreSQL persistence.
Key Decisions
Chose Flask over Django for faster iteration. Used Minimax TTS for natural-sounding multilingual speech. Implemented session-based learning to track progress without requiring accounts. Prioritized conversation quality over feature breadth.
The Problem
Cloud-based voice assistants raise privacy concerns and incur recurring costs. Users who want AI assistance with sensitive documents need an offline-first solution that never sends data externally.
The Approach
Built a fully offline voice assistant using Vosk for speech recognition and XTTS-v2 for neural text-to-speech. The entire pipeline runs locally with no cloud dependency.
The Result
Privacy-first assistant achieving 90%+ recognition accuracy with zero recurring costs. Handles document analysis and summarization entirely on-device.
Key Decisions
Chose Vosk over Whisper for lower resource requirements. XTTS-v2 selected for its natural-sounding multilingual output. Designed for consumer hardware without GPU requirements.
The Problem
Recruiters spend hours on repetitive screening tasks that could be automated. Manual candidate evaluation is inconsistent and prone to bias.
The Approach
Building a custom LLM assistant with RAG pipeline for automated candidate scoring, interview preparation, and data-driven hiring decision support.
The Result
In development. Targeting 70% reduction in initial screening time with more consistent evaluation criteria.
Key Decisions
Using LangChain for document processing pipeline. RAG architecture for context-aware candidate evaluation. Designed to augment recruiter judgment, not replace it.
Luis Zermeno
I started in recruitment — understanding people, roles, and what makes a great match. Then I discovered that AI could amplify that same intuition at scale. So I made the leap.
Now I build AI applications end to end: from problem definition through architecture, development, and deployment. I care about products that work for real users, not just impressive demos.