devLap

From My GitHub to Your Project — Meet the Developer Behind JARVIS OS, Lumen, Whisper Flow & TradeOracle

python dev.to

👋 Hi, I'm Franck Delmas

Full-Stack AI Engineer & System Architect

Building JARVIS OS — a distributed AI operating system with 928 autonomous agents running across 6 GPUs with 99.7% uptime.

Based in Toulouse, France.

🚀 My Projects

1. JARVIS OS — Distributed AI Operating System

The flagship project. A fully custom Linux-based OS designed to orchestrate 928 autonomous AI agents simultaneously.

  • 928 agents with specialized roles (voice, trading, automation, code review)
  • 6 GPUs (RTX 3080, RTX 2060, 4x GTX 1660S) = 46GB VRAM
  • 99.7% uptime on bare metal
  • Voice AI <300ms (Whisper large-v3 + Porcupine + EasySpeak)
  • 4 inference backends (LM Studio, Ollama, vLLM, Whisper CUDA)
  • 426K self-improve entries tracked via PostgreSQL
  • 56 skills across 32 systemd services

Tech: Python, CUDA, Docker, Bash, PostgreSQL, Redis, Flask

GitHub: https://github.com/Turbo31150/jarvis-linux

2. Lumen — Multilingual Transcription UI

Real-time transcription for 50+ languages. A beautiful React + TypeScript interface for live audio transcription.

  • 50+ supported languages
  • Real-time streaming transcription
  • Clean, responsive UI
  • Voice recording & playback
  • Export to text, SRT, JSON

Tech: React, TypeScript, Web Audio API, WebSocket

GitHub: https://github.com/Turbo31150/lumen

3. Whisper Flow — Real-Time CUDA STT Pipeline

Sub-280ms speech-to-text. A highly optimized CUDA-based pipeline for real-time speech recognition.

  • <280ms latency from speech to text
  • Whisper large-v3 on CUDA
  • Streaming support
  • Low VRAM footprint
  • Perfect for voice AI applications

Tech: Python, Whisper CUDA, CUDA kernels, PyTorch

GitHub: https://github.com/Turbo31150/jarvis-whisper-flow

4. TradeOracle — AI Trading with Multi-Model Consensus

Crypto trading automation with AI consensus. Multi-model trading decisions backed by technical analysis.

  • Multi-model consensus (Claude, Gemini, local LLMs)
  • CCXT integration (multi-exchange support)
  • Technical analysis engine
  • Risk management & position sizing
  • Backtesting framework
  • Integration with MEXC and other exchanges

Tech: Python, CCXT, Gemini API, TA-Lib, PostgreSQL

GitHub: https://github.com/Turbo31150/TradeOracle

5. BrowserOS — Multi-Browser Orchestration with MCP

AI agents that control browsers. Multi-browser automation using the Model Context Protocol.

  • Multi-browser orchestration (Chrome, Firefox, Edge)
  • MCP connectors for AI agent control
  • Headless + visible modes
  • Custom session management
  • Web scraping & automation workflows

Tech: Python, Puppeteer, Selenium, MCP protocol

GitHub: https://github.com/Turbo31150

6. ai-productivity-partner & ai-productivity-buddy

AI-powered productivity tools. TypeScript-based tools that integrate AI into daily workflows.

  • Task management with AI suggestions
  • Natural language commands
  • Integration with calendar & email
  • Custom AI agent plugins

Tech: TypeScript, Node.js, AI APIs

GitHub: https://github.com/Turbo31150/ai-productivity-partner

📊 GitHub Stats

  • 89 repositories
  • 45 stars
  • 621 contributions in the last year
  • 621 commits in April 2026 alone across 4 repos
  • Active contributor to: ai-productivity-partner, Soorce_facture, jarvis-linux
  • Recently created: whatsapp-prospection-ia-autonome, JARVIS-OMEGA, claude-code-stats

👉 https://github.com/Turbo31150

💼 Freelance Services

Based in Toulouse, France. Available for remote or on-site projects.

What I offer:

  • AI Agent Architecture & Deployment — Design and deploy autonomous agent systems
  • Distributed System Design — Multi-GPU clusters, Docker Swarm, load balancing
  • Linux Optimization & GPU Clustering — NVIDIA driver optimization, VRAM partitioning
  • Automation & Workflow Engineering — N8N, Bash scripting, API integrations
  • Voice AI Integration — Whisper, Porcupine, TTS pipelines
  • Trading Automation — Crypto trading bots, technical analysis, multi-exchange

Rate: 55€/h

Where to find me:

🛠️ Technical Stack

Category Technologies
AI/ML LM Studio, Ollama, vLLM, Whisper CUDA, Flash Attention
Languages Python, TypeScript, Bash, C#
Infra Docker Swarm, systemd, Redis, SQLite (WAL+mmap)
GPU NVIDIA RTX 3080, RTX 2060, 4x GTX 1660S
Voice Whisper large-v3, Porcupine, EasySpeak TTS
Trading CCXT, TA-Lib, MEXC, custom TA engine
Web React, TypeScript, Flask, N8N
Databases PostgreSQL, Redis, SQLite
Automation N8N, Bash, Python, Webhooks

🏆 Achievements

  • Hackathon Airia 2026 — Top participant
  • JARVIS OS — 928 agents, 99.7% uptime, 6 GPUs
  • Voice AI pipeline — Sub-300ms end-to-end latency
  • Trading automation — Multi-model consensus trading bot
  • Multi-GPU cluster — 46GB VRAM across 6 NVIDIA GPUs

📬 Let's Connect

Whether you have a project in mind, want to collaborate on open source, or just want to discuss distributed AI architecture — I'm always happy to talk.

Find me everywhere:

Platform Link
GitHub https://github.com/Turbo31150
LinkedIn https://linkedin.com/in/franck-hlb-80bb231b1
Codeur.com PRO member — https://www.codeur.com/
DEV.to https://dev.to/franckhlb_dev

Rate: 55€/h | Based in Toulouse, France

Thanks for reading. If you're building something similar or want to collaborate, reach out. Always happy to discuss distributed AI architecture.

— Franck Delmas | Full-Stack AI Engineer & System Architect | JARVIS OS Creator

Source: dev.to

arrow_back Back to Tutorials