High-Performance Architecture & AI Integration Specialist

🚀 From a passionate self-taught developer to a contributor in major AI projects. I have been breaking barriers in web and software development for over 4 years, evolving from the classic MERN stack towards complex low-latency architectures, Edge systems, and AI Agents.

A dream come true! I have officially participated in the development of the QwenLM ecosystem, one of the most powerful LLMs in the world.

• The Context: I contributed to qwen-code (Qwen-Agent). It is worth noting that qwen-cli is architecturally inspired by Google's gemini-cli, demonstrating that I am working with code standards adopted by major tech giants.
• The Contribution: I implemented critical logic for the Model Context Protocol (MCP) integration, fixing gaps in the agent instructions.
• The Background: This wasn't just a copy-paste job. I achieved this by deep-diving into advanced Agentic AI and MCP concepts (special dedication to my self-study marathons with AI Anytime courses) to truly understand how to give LLMs "superpowers" via tools.

🔗 See my accepted Pull Request: Advanced Logic for MCP Integration

I am building an experimental next-gen browser extension with integrated ASR (Automatic Speech Recognition) and TTS (Text-to-Speech) to free developers from the keyboard and prioritize physical health.

• Architecture: Backend running on Bun with native WebSockets to achieve sub-millisecond latency.
• Extreme Performance: Implementation of memory piping with 16kb buffers to handle real-time audio streaming without touching the disk, effectively avoiding Garbage Collector blocking.
• The Challenge: Overcoming Manifest V3 limitations by utilizing Offscreen documents for persistent and smooth audio processing.

I don't limit myself to a single framework; I choose the tool that makes the code fly.

• Core: JavaScript (ESNext), TypeScript, Bun (Current favorite for speed).
• Frontend: Next.js, React, Astro (for static/content-heavy sites), TailwindCSS.
• Backend & Runtime: Node.js, Bun (Native WebSockets), Express, Trpc.
• Data & State: MongoDB, PostgreSQL (Prisma), In-Memory File Systems (Streams/Buffers).
• AI & ML Engineering: Integration of LLMs (Gemini, Qwen), MCP implementation, Whisper (ASR), TTS pipelines, WebGL/Three.js visualizations.
• Mobile: React Native (Android/iOS).

I started 4 years ago studying day and night (kudos to the community: Midu, Fazt). My philosophy is Mobile First and extreme optimization. I have gone from building e-commerce layouts to designing systems that process real-time audio and collaborating on the codebase of global AI models.

"I don't just want it to work, I want it to fly. If I have to go down to the level of buffers and streams to optimize, that's where you'll find me."

I am open to collaborations on Open Source projects, AI integrations (especially Agents and MCP), or complex web architecture challenges.

• 📧 Email: songokumbl@gmail.com
• 💻 GitHub: sontoriyama