Enterprise AI.
Running Safely
Deploy strict, air-gapped AI engines that never send data to the cloud. Complete data privacy. Zero API fees. 100% Offline.
What Is This Project?
Vox Amelior is a real-time enterprise transcription and analytics platform that runs entirely offline on consumer hardware. It performs live speech-to-text, speaker identification, sentiment analysis, and 36 specialized ML engines - all on a single GPU with just 6GB VRAM. No cloud dependencies. No API fees. Complete data privacy. See it in action.
What this project demonstrates about my engineering capabilities:
332K+ lines across 7 coding languages and 12 microservices. Not a tutorial project, a production system.
Custom semaphore system runs LLM, ASR, and ML models on a single consumer GPU through intelligent scheduling.
Titan AutoML, causal inference, symbolic regression, time series. Specialized engines for enterprise analytics.
AES-256 encryption, JWT auth, Docker secrets, network isolation. Zero data exfiltration by design.
Click Any Technology to See the Code
Every technology listed here links directly to its implementation in the repository.
Core AI/ML
ML Engines
Backend
Infrastructure
Security
Architecture Patterns
Featured Work
Production systems designed for real-world deployment
NeMo AI Platform
A comprehensive multimodal AI platform featuring real-time speech transcription, speaker verification, emotion analysis, and RAG-powered conversational AI. All orchestrated on a single consumer GPU through intelligent resource management.
View Live Demo Explore ArchitectureGPU Coordination
Custom semaphore-based system for sharing a single GPU across multiple ML models with preemptive scheduling.
Privacy-First
100% local processing with AES-256 encrypted databases. No cloud dependencies, no data leaves the device.
Production Ready
Full Kubernetes deployment with health checks, graceful shutdowns, and horizontal scaling support.